toybox/www/news.html
<<
>>
Prefs
   1<html><head><title>toybox news</title>
   2<!--#include file="header.html" -->
   3
   4<p>Toybox combines common Linux command line utilities together
   5into a single BSD-licensed executable that's simple, small, fast,
   6reasonably standards-compliant, and powerful enough to turn Android into
   7a development environment. See the links on the left for details.</p>
   8
   9<h2>News</h2>
  10
  11<a name="12-10-2017" /><a href="#03-10-2017"><hr><h2><b>October 12, 2017</b></h2></a>
  12<blockquote><p>Would it save you a lot of time if I just gave up and went
  13mad now?</p>
  14<p> - The Hitchhiker's Guide to the Galaxy</p>
  15</blockquote>
  16
  17<p><a href=downloads/toybox-0.7.5.tar.gz>Toybox 0.7.5</a>
  18(<a href=https://github.com/landley/toybox/releases/tag/0.7.5>git commit</a>)
  19is out.</p>
  20
  21<p>This is another <b>mostly bugfix release</b>. I delayed it a bit trying to get some
  22of the many half-finished projects (dd, ping, lsof, iconv, cut, sysconf,
  23the cp --parents option...) finished/promoted/tested, but decided adding big
  24things at the end of the dev cycle would be too destabilizing.</p>
  25
  26<p>A couple <b>new features</b> slipped in anyway. Ilya Kuzmich added head -c
  27and strings -t (and corrected strings' output in a couple places). Elliott
  28Hughes taught file to recognize Macintosh (Mach-O) binaries.</p>
  29
  30<p><b>NOTE</b>: the chrt command is broken when built against musl-libc
  31because that project's maintainer decided he didn't like the system calls
  32it depends on, so he
  33<a href=https://git.musl-libc.org/cgit/musl/commit/?id=1e21e78bf7a5>removed
  34them from his libc</a>. The prebuilt binaries are built against musl,
  35so the chrt command there will always say -ENOSYS. (It works fine built against
  36glibc, and presumably bionic or uClibc.)</p>
  37
  38<p><b>Bugfixes</b>: Remounting (mount -o remount,rw) should work properly now,
  39and we fixed another place ps was segfaulting when /proc
  40entries vanished out from under us (a hard to hit race condition
  41mostly noticeable in long-running "top" exiting; error value was treated
  42as a pointer). Elliott Hughes fixed a -Wformat warning in expr.c,
  43and made xargs split the command
  44line at ARG_MAX (which is 1/4 the stack size ulimit, which normal users
  45can arbitrarily increase but execve() will complain if they don't).
  46Josh Gao made netcat -l exit after handling a request, and reported a
  47seq bug (seq "1000000 10000001" output 1e+06, fixed now). Zach Riggle fixed a
  48broken URL in the README. Rob fixed env -0 checking the flag wrong so
  49it didn't work in combination with other flags (typo).</p>
  50
  51<p><b>Plumbing</b>:
  52Replaced mbrtowc() with a new utf8towc() that doesn't have a context
  53struct or care about locale. A bugfix in comma_scan() (wasn't removing an entry at the end
  54of the list) is why remount _sometimes_ worked (depending on argument
  55order), xgetaddrinfo() is now separate from xconnect(),
  56atolx_range() learned the w (word=2) suffix, and b is now (block=512) instead
  57of (byte=1).</p>
  58
  59<p><b>Build</b>: Patrick Oppenlander added a workaround for a bug in config2help.c
  60that resulted in segfaults on newer toolchains. (It was actually a use
  61after free error; the build infrastructure isn't nearly as heavily audited as
  62code that gets installed on the target; oops.) A new GITHASH
  63environment variable can force the build version from
  64the command line, and the build checks for .git in the top directory
  65before asking git what our version is (so it isn't confused by ../../.git
  66in an enclosing directory). Added a build #warning about musl intentionally
  67breaking chrt (as with nommu fork() musl now provides a broken
  68stub function so compile-time probes for its existence think it's there, and
  69you can't run the result to test behavior when cross compiling).</p>
  70
  71<a name="19-06-2017" /><a href="#19-06-2017"><hr><h2><b>June 19, 2017</b></h2></a>
  72<blockquote><p>It is a well-known fact that those people who most want to rule people are, ipso facto, those least suited to do it.
  73To summarize the summary: anyone who is capable of getting themselves made President should on no account be allowed to do the job.</p>
  74<p>- The Hitchhiker's Guide to the Galaxy</p>
  75</blockquote>
  76
  77<p><a href=downloads/toybox-0.7.4.tar.gz>Toybox 0.7.4</a>
  78(<a href=https://github.com/landley/toybox/releases/tag/0.7.4>git commit</a>)
  79is out. No new commands this time, but
  80<b>chrt</b> and <b>dmesg</b> got promoted out of pending.</p>
  81
  82<p><u>New features</u>:
  83Rob rewrote paste, which should work much better now, and added grep
  84-M and -S to match and skip wildcards respectively (useful with -r).
  85Elliott's updated dmesg has -T and --color. The file
  86command can recognize gzip now, uptime grew -s, date grew %N, env knows - as a first argument
  87means -i (posix!) and grew -0, ls defaults to -b
  88instead of -q now when there's a tty, and ls has a new -ll option (with
  89--full-time as a compatibility synonym) showing nanoseconds and (for some
  90reason) timezone. (Why do individual files have timezones?) Elliott added
  91"uudecode -o -" support. Illya Kuzmich taught head -v and -q. The cpio
  92code no longer adds the "TRAILER!!!" entry by default (initramfs extractor
  93doesn't care) without which you can concatenate cpio archives with "cat".
  94(Use the new --trailer option if you want the legacy behavior.)
  95In pending, fdisk compiles now and tar understands bzip2.</p>
  96
  97<p><u>Build</u>:
  98The "make install_airlock" target now symlinks bc from the host because
  99the kernel <a href=https://landley.net/notes-2013.html#28-03-2013>inexplicably</a>
 100needs that to build. This was motivated by
 101<a href=https://github.com/landley/mkroot>mkroot</a>, which builds under
 102a toybox airlock directory.</p>
 103
 104<p>Lots of work on the test suite, mostly from the Android guys who are now
 105running it under Android. This fixed several existing tests that didn't
 106pass, made more tests run on a toybox-only system, and so on. The test suite
 107infrastructure now has a second testing function,
 108"testcmd", which supplies the command name being tested (bypassing
 109shell builtins).</p>
 110
 111<p>Various android build and config fixes, getting closer to being able
 112to let android someday use scripts/make.sh instead of generated/* snapshots.
 113Also more work into building under android's NDK; not quite there yet
 114but much closer.
 115Use nproc in scripts/make.sh detect available processors (so you can control the SMP level with taskset).
 116Removed the old uClibc compatibility glue, it's been 5 years since their
 117<a href=http://lists.busybox.net/pipermail/buildroot/2016-December/180102.html>last release</a>.</p>
 118
 119<p>The new config option TOYBOX_PEDANTIC_ARGS checks arguments when there
 120are no arguments, so things like "uptime" no longer silently ignore arguments
 121you pass but instead refuse to run.</p>
 122
 123<p><u>Docs</u>:
 124The FAQ now has more than one entry. Commands no longer output the full
 125help text for argument errors but instead just say "See %s --help" with the
 126command name (in addition to the actual error message).
 127Elliott did a big period-ectomy on all the --help text, and
 128we cleaned up some tab/space inconsistency. The
 129non-html help -a output now has separators with the command name.
 130The top/iotop and pkill/pgrep help text now describe a lot more of what
 131the commands can do. Twitter's code of conduct page went down so we
 132mirrored the text locally.</p>
 133
 134<p><u>Bugfixes</u>:
 135Fixed a race condition in ps/top where a process that exited right as we
 136read its data returned a different error value than we were expecting (which
 137was causing long-running top instances to occasionally exit),
 138mount now gives an error if it can't autodetect the filesystem
 139type, ps no longer queries the terminal size when output isn't to a tty
 140(so "ps -A | cat" doesn't vary), date's chkmktime() was replaced with
 141simple range checks for fields (to avoid false positives from things like
 142timezones and daylight savings time), removed %s from date's help (we
 143didn't implement it, we have @seconds[.nanoseconds] instead), fixed
 144zcat's buffer flush logic (which was always failing on files larger
 145than 32k), and factor now detects requests for numbers >64 bits and fails
 146loudly instead of producing incorrect answers.
 147Elliott fixed touch -a/-m (they were backwards), and allowed ':' in
 148setprop's property names. Grep now exits with 2 for errors (so -q can
 149distinguish "didn't find" from "didn't work"), doesn't stop on symlinks
 150that point nowhere (there was an error_exit() that should just be a warning),
 151and provides error messages for files we could open but not read.</p>
 152
 153<p><u>Library</u>:
 154New library functions: strend() complements strstart(), minof()/maxof()
 155are min/max macros that evalute arguments once and autodetect type (why
 156isn't this in libc?), xmmap() checks MAP_FAILED (which is not NULL).</p>
 157
 158<a name="21-02-2017" /><a href="#21-02-2017"><hr><h2><b>February 21, 2017</b></h2></a>
 159<blockquote><p>Only six people in the Galaxy knew that the job of the
 160Galactic President was not to wield power but to attract attention
 161away from it. Zaphod Beeblebrox was amazingly good at his job.</p>
 162<p>- The Hitchhiker's Guide to the Galaxy</p></blockquote>
 163
 164<p>Despite everything, <a href=downloads/toybox-0.7.3.tar.gz>Toybox 0.7.3</a>
 165(<a href=https://github.com/landley/toybox/releases/tag/0.7.3>git commit</a>)
 166is out. The <u>new commands</u> this time are <b>ftpget</b>, <b>ftpput</b>, <b>microcom</b>, and <b>ascii</b>.<p>
 167
 168<p>We also had two command _demotions_ out of defconfig:
 169<b>hostid</b> got moved to toys/example and
 170switched to "default n" because despite <a href=http://pubs.opengroup.org/onlinepubs/9699919799/functions/gethostid.html>still being in posix</a>
 171the concept of a unique 32 bit number identifying a system is something
 172Linux outgrew about the time Pauline Middelink wrote the first IP
 173Masquerading code. And Elliott did a complete rewrite of <b>dmesg</b> introducing
 174two codepaths that I didn't get a chance to unify and didn't want to
 175hold up the release for, so that's back in pending.</p>
 176
 177<p><u>New features</u>: Rob added units to <b>find</b> -atime and friends
 178(with the legacy -amin alias). Elliott added color and -w to dmesg, fallocate
 179-o, and improved file's ELF parsing. Steve Muckle added -d and finit_module
 180support to modprobe. Rob and Elliott tweaked the
 181ps/top display format a bit more (extending the USER field from 8 to 18 chars
 182and putting + at the end of string fields that got truncated).
 183df -a isn't entirely new, but wasn't documented and needed a bugfix.</p>
 184
 185<p><u>Bugfixes</u>:
 186Last release broke oneit because -c didn't get moved to xopen_stdio() (oops).
 187Rob and Elliott simultaneously spotted ps padding each line to 99999
 188chars when there's no tty (serial console or adb); now it pads to 80 in
 189that case but also switches on -w to avoid field truncation. The "tty"
 190field also sometimes had trailing debris (that's fixed now). And "top" was
 191endlessly redrawing with out tty because receipt of the ANSI size probe
 192results would set SIGWINCH, and handling that sent another ansi probe. (Sigh.)
 193And while we're there, replace "ADDR" with "BIT" in ps -l so there are
 194more than 4 chars left for the "CMD" field on 64 bit systems.</p>
 195
 196<p>Izabera pointed out that split -b and -l can't mix, and suggested seq should
 197multiply to avoid accumulating rounding errors from repeated fractional
 198increments. Wang Xiao Jian fixed a bug in sort -k.
 199Elliott let getprop use the @ character in property names, and
 200Dimitry Ivanov removed the name length limit for system properties.
 201Elliott also improved some error reporting and improved top -H's display
 202of thread names.
 203
 204<p>Josh Gao pointed out that recursive operations on . and .. could be ignored
 205in chmod -R (and the resulting generic fix to dirtree_notdotdot() fixed
 206it in several other places).</p>
 207
 208<p>Justin Cormack caught tar producing a warning to stdout that screwed up
 209"tar c" to stdout.
 210Rob fixed an option parsing bug (where switching off a --longopt in menuconfig
 211confused the parser), and another one where an option excluding itself
 212(ala "abc[-ab][!abc]" with "command -a -b") would segfault.</p>
 213
 214<p>There's some sort of gcc stack over-optimization bug where musl-libc's
 215version of vfork() doesn't get marked with attribute(returns_twice) so
 216stack varabiles in the same function after that get semi-randomly overwritten
 217when the optimizer decides to reclaim the space. So add the attribute
 218to the function the XVFORK() wrapper macro calls. (It's a nommu thing.)</p>
 219
 220<p>Fixed a couple variable size mismatch bugs that were only tested on 64 bit
 221(printf %x 64) or only tested on 32 bit (modprobe), removed some
 222unnecessary casts in stat.</p>
 223
 224<p>Continuing attempts to build under Android NDK brought up that posix
 225defines the global 'stdout' as a macro, which older versions of bionic
 226turned into an array member, but a function was using it as an argument
 227name. (This worked in the AOSP build because it only builds against current
 228bionic, where there's a global 'stdout'.)</p>
 229
 230<p>Several commits argued with clang's warning generation, eventually
 231settling on a variant of __attribute__((__shut_up__)).</p>
 232
 233<p>Android should no longer give spurious error messages
 234when you "ps -A | head" about EPIPE on output. (Older versions of bionic
 235set an error handler on SIGPIPE, but it shouldn't do that now. More recent
 236versions of adb set the SIGPIPE handler to SIGIGN instead of SIGDFL,
 237leading to write returning an error message instead of silently killing
 238the program. So we set it back to the default.)</p>
 239
 240<p><u>Docs</u>:
 241Removed website link to the gmane archive (which didn't survive gmane's
 242change of ownership). The FAQ now answers a _second_ question. (Woo!)
 243Some roadmap updates.</p>
 244
 245<p><u>Build tweaks</u>:
 246Upgraded "make install_airlock" target to only warn about missing
 247commands (unless $PEDANTIC is set) when it sets up the hermetic build
 248path. (The plan is still to implement everything but the toolchain
 249binaries in toybox, but in the meantime we're symlinking other stuff from
 250the $HOST that isn't ready yet. See
 251<a href=https://github.com/landley/mkroot>mkroot</a> for an example using
 252this.)</p>
 253
 254<p>Elliott and Rob continue to poke at building toybox with Android's NDK,
 255but it's a work in progress (<a href=http://lists.landley.net/pipermail/toybox-landley.net/2016-December/008767.html>thread</a>). Various changes
 256removing libcutils dependencies and adding an selinux dependency to getprop
 257are fallout from this.</p>
 258
 259<p>Cross-compiling from Macs needs to use "gsed" instead of apple's
 260version, so teach the build to use that name if it exists in the $PATH.
 261If you try to build without running config first, you should get better
 262error reporting now. Added a workaround for Centos' broken "which" command
 263producing output when it _can't_ find a name in the $PATH.</p>
 264
 265<p><u>Library</u>:
 266The new dirtree flag DIRTREE_PROC skips non-numeric entries so things
 267like ps and top can scan /proc more efficiently.</p>
 268
 269<a name="21-10-2016" /><a href="#21-10-2016"><hr><h2><b>October 21, 2016</b></h2></a>
 270<blockquote><p>Probability factor of one to one. We have normality. I repeat,
 271we have normality. Anything you still can't cope with is therefore your
 272own problem.</p><p> - The Hitchhiker's Guide to the Galaxy</p></blockquote>
 273
 274<p><a href=downloads/toybox-0.7.2.tar.gz>Toybox 0.7.2</a>
 275(<a href=https://github.com/landley/toybox/releases/tag/0.7.2>git commit</a>)
 276is out.</p>
 277
 278<p>During this development cycle, Elliott Hughes <a href=http://androidbackstage.blogspot.com/2016/07/episode-53-adb-on-adb.html>got interviewed
 279on the ADB podcast</a>
 280and Rob Landley <a href=http://linuxluddites.com/shows/episode-88/>got interviewed on Linux Luddites</a> (<a href=http://linuxluddites.com/shows/episode-11/>again</a>).
 281Both talk about toybox and many other things.
 282The web page also grew a new <a href=faq.html>FAQ page</a>, currently with
 283just the one.</p>
 284
 285<p>New comands added to defconfig are <b>tunctl</b>,
 286<b>log</b>, <b>start</b>, <b>stop</b>, and <b>sendevent</b>.
 287The commands <b>file</b> and <b>netstat</b> got promoted out of pending.
 288Pending added <b>chrt</b>, <b>setfattr</b>, and <b>getfattr</b>, and saw
 289a lot of cleanups to diffstat and dd but not enough to promote them to
 290defconfig yet. A new toys/net directory was added, moving ifconfig, netcat,
 291netstat, rfkill, and tunctl there so far.</p>
 292
 293<p><b>Upgrades</b>: All commands now parse --version when they understand
 294--help, but "true" and "false" should now ignore their arguments entirely.
 295We taught stat to handle "%12x" and "%.12x" printf-style escapes, which
 296apparently other versions do. The ifconfig output now shows the interface's
 297device driver. Added patch -d and --dry-run, wc can now do -cm together,
 298find has a NOP -noleaf so scripts that use that don't break, add -c to md5sum
 299and sha1sum. Elliott taught ps to treat extra aguments as additional -p
 300pids, implemented xxd -s, did a number of upgrades to file (added -HL,
 301support for ar files, improved ELF support to report android API level
 302and stripped/not stripped and it no longer prints a guessed build ID type).
 303Elliott also added optional build-time support for using openssl's
 304assembly-optimized md5sum/sha1sum implementations (leading to a new
 305<a href=design.html>design</a> policy on shared libraries).</p>
 306
 307<p><b>Bugfixes</b>: Too many fixes to "ps" and "touch" to list, from both
 308Elliott and Rob. Rob taught sed to handle s/[[:space:]/]// type sequences
 309properly, switched grep to a better
 310workaround for <a href=https://sourceware.org/bugzilla/show_bug.cgi?id=17829>glibc bug 17829</a>, made sed -i preserve ownership when run as root,
 311made du max out at 2 terabytes instead of 2 gigabytes on 32-bit systems
 312(it was always designed to, but was missing a typecast),
 313fixed the option parsing infrastructure (config options that remove command
 314line options got the placeholders wrong), fix to printf for printing
 315octal digits and handling the (posix-mandated) difference between %b and
 316non-%b octal output, reading from "-" no longer closes stdin when done,
 317netcat -L works with nommu (although it may need more portability work),
 318and you can now "make test_scankey" if you want to. Several commands
 319(stat, makedeves, chgrp, cp, find) handled user name lookup failure badly
 320(stat was segfaulting if you interrogated a file belonging to a nonexistent
 321user, "chown 12345 file" errored out if you didn't have that user
 322in /etc/passwd... now they should all print/accept the number when
 323appropriate). "LC_ALL=C ls -Cs --color" produces the same output
 324as other versions (two spaces padding, -k hardwired on).<p>
 325
 326<p>Kyungsik Lee fixed a bug
 327in cp (readlink() doesn't actually null terminate the string it reads in),
 328Elliott Hughes made pgrep/pkill return success/failure, fixed trailing
 329whitespace in netstat, fixed a SMACK symbol conflict due to linux/xattr.h
 330changing, fixed ls -sh, and added a lot of
 331stuff to the <a href=roadmap.html>roadmap page</a>.
 332Izabera pointed out cmp -l and -s can't be selected at the same time,
 333that timeout was never actually checking -v, that ls should default to -q
 334when output is to a tty, and that "file -" would sometimes try to open "-"
 335instead of stdin.
 336Usischev Yury pointed out a use after free error, and that id shouldn't
 337call exit() directly. Matthias Urhahn pointed out that stat(2) returns
 338hardwired 512-byte units, so stat.c was wrong. David Hedges pointed out
 339that route could only handle 10 character interface names when the kernel
 340can do 15 (it's still in pending for a reason, but fixed). Evgenii
 341Stepanov found and helped diagnose one of the more subtle ps bugs fixed
 342this time around.</p>
 343
 344<p>Calling "make test_blah" no longer causes make to error out if the last
 345test fails (and thus returns a nonzero error code).
 346Building single commands and the multiplexer used to require a "make clean"
 347between them (because they had different config files both of which were older
 348than generated/config.h so it didn't get rebuilt; now it just always
 349rebuilds it).</p>
 350
 351<p>The defconfig build is now slightly less broken on older centos versions
 352(although <a href=http://lists.landley.net/pipermail/toybox-landley.net/2016-September/008664.html>the consensus</a> is that Centos is just generally broken).<p>
 353
 354<p>Several commands were over-using xprintf(), which flushes its output
 355to check for error (something you only need to do maybe once per line,
 356and even then maybe only in loops because xexit() flushes and checks
 357ferror() for you and adjusts the exit code if we wrote stuff to stdout
 358that couldn't be printed). Lots of little flushes are inefficient,
 359so most things can use normal printf(). (Retransmission
 360of short writes is presumably libc's problem since it's buffering the
 361output and all.)</p>
 362
 363<p><b>Library:</b>
 364New library functions readlink0() and readlinkat0() which properly null
 365terminates the symlink value (which the stock libc function inexplicably
 366doesn't).
 367The new do_lines() function interates reading lines from a filehandle
 368and calling a function on each line.
 369New function pollinate() factoring out netcat's poll() loop so things
 370like telnet can use it.
 371New functions getusername() and getgroupname() return a
 372char * given a uid/gid (and return a string representation of the number
 373if the lookup fails), and xgetpwnamid/xgetgrnamid were renamed to
 374xgetuid/xgetgid and now return an integer instead of a struct (also helping
 375handle lookup failures, you can still return the uid/get for "12345").</p>
 376
 377<p>Switched atolx() to use long long internally.
 378Renamed xopen() to xopen_stdio() and made a new xopen() that never returns
 379stdin, stdout, or stderr (duping /dev/null into the filehandles as necessary).
 380New function xopenro() opens a file read only with one less argument, and
 381understands that "-" means stdin. New flag WARN_ONLY tells these functions
 382to just print a warning on failure, and return -1 instead of exiting.
 383Misc new functions like openro() which defaults to the WARN_ONLY behavior
 384and notstdio() which dup()s a filehandle up beyond stdin/out/err backfilling
 385with /dev/null as necessary. The WARN_ONLY flag let us remove the failok
 386argument from loopfiles().</p>
 387
 388<p>New TOYFLAG_NOHELP disables --help processing (which "true" and "false"
 389should not do).</b>
 390
 391<p>The test suite now has NOSPACE=1 to ignore whitespace (using diff -b to
 392check results), which helps TEST_HOST pass the same tests as toybox.
 393Fixes to chattr and date tests. It also has a new variable $C with the
 394absolute path to the command being tested (bypassing shell builtins),
 395and a function testcmd() which is just like testing() except it prepends
 396the command name ($C) to the test command line as well as the test
 397description.</p>
 398
 399<a name="02-06-2016" /><a href="#02-06-2016"><hr><h2><b>June 2, 2016</b></h2></a>
 400<blockquote><p>When the 'Drink' button is pressed it makes an instant but
 401highly detailed examination of the subject's taste buds, a spectroscopic
 402analysis of the subject's metabolism, and then sends tiny experimental
 403signals down the neural pathways to the taste centres of the subject's
 404brain to see what is likely to be well received. However, no-one knows
 405quite why it does this because it then invariably delivers a cupful of
 406liquid that is almost, but not quite, entirely unlike tea.</p>
 407<p> - The Hitchhiker's Guide to the Galaxy</p></blockquote>
 408
 409<p><a href=downloads/toybox-0.7.1.tar.gz>Toybox 0.7.1</a>
 410(<a href=https://github.com/landley/toybox/releases/tag/0.7.1>git commit</a>)
 411is out. (Yes, I forgot to update the --version string, but I already
 412uploaded the <a href=downloads/binaries/0.7.1>binaries</a>.)</p>
 413
 414<p>The website has https support now, you can "make cat ps ls"
 415to get standalone commands (and "make list list_pending" to see what's
 416available), and a whole lot of bugfixes and new options to existing
 417commands.</p>
 418
 419<h2>New Commands</h2>
 420<p>Rob implemented <b>ulimit</b>. In pending, Elliott Hughes implemented
 421file. and Lipi Lee implemented a simple wget. (Pending also had minor
 422cleanups to more and lsof, but no promotions this time around.)</p>
 423
 424<h2>New Options</h2>
 425<p>Izabera implemented env -u, suggested adding seq -w, made factor
 426use full unsigned 64 bit math even on 32 bit platforms, pointed out base64
 427-w0 should disable wrapping, and sped up wc -c.
 428Elliott Hughes added mount -o relatime, xxd -p -r, and od -w.
 429Sameer Pradhan (or possibly Bilal Qureshi) suggested adding stat -tL -c %m%t%T.
 430Tom Cherry added getprop -Z. Paul Barker added hostname -b and -F.
 431Rob added ls -b, made ls -q work with utf8,
 432made sed -f - read from stdin, and added top -O (like ps -O).</p>
 433
 434<p>Elliott and Rob <b>added Thread support to ps and top</b>,
 435with -o TID, TNAME. We also added -o PCY (android scheduling policy),
 436-o BIT (process is 32 or 64 bit), and -o TNAME now shows the parent
 437command name for threads.</p>
 438
 439<h2>Documentation</h2>
 440
 441<p>Rob added the sed invocations to convert tabs/spaces and back to
 442design.html. Isaac Dunham updated hexedit's help text.
 443Jakob Flierl pointed out a broken URL in the README.</p>
 444
 445<p>Rob also redid the naming scheme of sed's pattern manipulation code to
 446remove the gratuitous references to Roger Zelazny's "Amber" series, since it
 447was confusing people.</p>
 448
 449<h2>Bugfixes</h2>
 450
 451<p>Grep -H and -n should now work properly with -ABC. Andy Chu pointed out an
 452out of bounds access for zero length lines in rev, fixed a buffer overflow
 453in diff -r, and fixed operator precedence in expr (although Rob is rewriting
 454chunks of expr so toysh can use its plumbing for $(( )) ).
 455Patrick Ohly fixed the too-aggressive suid permission dropping logic.
 456Josh Gao fixed a segfault when find -iname got no argument, and
 457made tail -f work right with just one file argument.
 458Tom Marshall cleaned up tar's long filename support and improved
 459the tar tests, and reported another find bug (with -iname -o -iname not
 460tracking copy lifetimes properly) that got fixed.</p>
 461
 462<p>Elliott Hughes fixed wc -c to not trust zero length files to actually be
 463zero length (/proc does that), fixed "mount -o rw,remount /system" on
 464Android, removed trailing spaces on ps -o cmdline, fixed pkill -9 and
 465the corresponding tests, made "insmod -" work, fixed top -b and tail -NUM,
 466pointed out that ps shouldn't trim numeric fields for display size limits,
 467and added some more
 468explicit "sort" calls to make pipelines so build tempfiles are easier to cache.
 469Rob <a href=https://github.com/landley/toybox/commit/32b3587af261>fixed an insane sed thing</a> the perl 5.22 build was doing.
 470Fixed mount -o to properly pass in leftover string data, and
 471<a href=http://lists.landley.net/pipermail/toybox-landley.net/2016-March/004790.html>documented how to use toybox to mount nfs</a> (warning: kernel patch
 472to fix some bitrot in the kernel NFS driver's string parsing attached
 473to that message).</p>
 474
 475<p>George Burgess IV corrected some variable types in traceroute.
 476Base64 now wraps == properly. Fixed two bzcat segfaults reported by
 477John Regehr. Andy Chu found a segfault in "sed -e 'c\'" with no trailing
 478line, and implemented mv -n and cp -n. The cyanogenmod guys pointed out that
 479cp -a shouldn't complain if a non-root user can't chown, and we added
 480the output path to cp -r error messages while we were there (before was just
 481the filename).</p>
 482
 483<p>Samuel Holland fixed blkid's handling of vfat labels, and
 484fixed a segfault when basename was passed an empty string and an empty
 485suffix. Davis Mosenkovs fixed touch -t seconds parsing.
 486Rob fixed a bunzip bug reported by John Regehr (the bad CRC
 487error message was printing a NUL argument).</p>
 488
 489<p>Not all of the commands build standalone, but more of them do now;
 490scripts/single.sh can now build a "mv" that isn't actually "cp".
 491The dependencies are more granualr, so "make top; make ps" no longer
 492produces a broken ps that ignores -A (because ps.o wasn't getting rebuilt
 493even though top had the FLAG macros for -A zeroed).</p>
 494
 495<h2>Build</h2>
 496<p>Rob added a <b>warning when building commands out of
 497pending</b>. (The pending directory is full of code that hasn't been
 498properly vetted. Use at your own risk.)</p>
 499
 500<p><b>New build targets let you build individual commands by name</b>, ala
 501"make ls cat ps", and you can run the test suite for each standalone
 502command with "make test_ls" and such.
 503"make list" shows all such standalone commands in defconfig, and
 504"make list_pending" shows unfinished commands from toys/pending
 505("make list list_pending" shows both together). "make clean" now deletes
 506these filenames at the top level, and the corresponding unstripped files
 507live in the directory generated/unstripped.</p>
 508
 509<p>Nicholas Boichat suggested switching make.sh to use $! for process
 510enumeration during parallel builds (which is both more efficient and more
 511portable), and suggested shell builtin replacements for wc/awk/sed so the
 512build loop has fewer forks now.</p>
 513
 514<p>Lots of work on the test suite, much of it due to Andy Chu. It now
 515consistently prints the name of the command being tested at the start of each
 516test (and the common infrastructure does that, not each individual test), and
 517"make tests" actually runs all the available tests now.
 518Seperated pgrep and pkill tests, split lsattr/chattr, added fstype and base64
 519tests. The "tests/files" directory now collects files for tests to
 520use, with blkid, bzcat, and utf8 subdirectories: the $FILES variable
 521gives a path to it, so "$FILES/blkid/ext2.bz2" and so on.
 522The testsuite now has test files with 3 different types of "not utf8 output"
 523sequences that require escaping, plus some combining character torture
 524tests, direction reversals, and so on.</p>
 525
 526<p>Added dependencies on TOYBOX_FORK to various pending commands that need
 527nommu conversion (which should fix the allyesconfig build).</p>
 528
 529<p>Static builds with selinux should work again.</p>
 530
 531<h2>Library</h2>
 532
 533<p>New bufgetgrgid() and bufgetpwuid() functions cache previous lookup info
 534rather than repeatedly traversing /etc/passwd and /etc/group (which is slow).
 535Added xpipe() to lib to catch pipe creation failure.
 536The HELP_ macros generated by config2help.c now use a capital prefix
 537to avoid collicing with help_exit() and such.</p>
 538
 539<p>The dirtree infrastructure got a cleanup pass in preparation for adding
 540infinite recursion depth support (needed by rm -r), updated the
 541<a href=code.html#ib_dirtree>documentation</a> to describe the new
 542semantics (removing dirtree_start() and adding dirtree_flagread()).
 543Now dirtree_recurse() takes the new dirfd as an argument.</p>
 544
 545<p>Split out _xexit() from xexit() and let sigatexit() set multiple
 546callbacks.</p>
 547
 548<p>For years the man pages have said to #include <sys/types.h> to get
 549major/minor/makedev but now that glibc
 550<a href=https://sourceware.org/ml/libc-alpha/2015-11/msg00253.html>has vowed
 551to break existing programs</a> and replace it with another nonstandard header
 552not in posix or lsb, we added our own functions to lib/ to do the transform
 553ourselves (based on what the kernel actually expects).</p>
 554
 555<h2>Portability</h2>
 556
 557<p>Debian unstable started needing an extra header #include for some reason, and
 558although printf("%.*s", INT_MAX, s) worked fine on Ubuntu 12.04 it
 559didn't on 14.04, so added a workaround for that. Typecast a printf because
 560wchar_t isn't a rigidly defined size. RLIMIT_RTTIME was
 561added to the kernel in 2008 but you can't expect uClibc to have noticed yet,
 562nor did it #define MS_RELATIME (added in 2006), or prlimit (2010)...
 563(Given the improvements in musl and bionic, uClibc support may be dropped
 564in a future release.) Given that the xattr functions were added during
 565linux 2.5, we can #include its header unconditionally.</p>
 566
 567<p>Renamed basename_r() to something else to avoid conflicting with freebsd's
 568libc, and both scripts/install.h and scripts/config2help.c no longer include
 569toys.h (to make cross-compiling from systems we don't run on easier).</p>
 570
 571<p>Debian bug 635570 did something unspeakably nonportable, depending on
 572"sed -e 'a\'" (with no next line of the pattern, so an unterminated
 573continuation) to add a newline to the last line of input if and only if
 574that last line of the input didn't have a newline, and to take no other
 575action. This is well into "depending on a bug" territory, but we implemented
 576it because otherwise Debian's install broke. (Of course this behavior
 577is undocumented, non-obvious, and doesn't really make logical sense.)</p>
 578
 579<p>CONFIG_TOYBOX_NORECURSE now disables the stack measuring logic (which
 580was giving some "security" code fits). Also we typecast pointers to (long)
 581before comparing them to avoid spurious compiler "optimizations" that
 582break the code.</p>
 583
 584<a name="02-02-2016" /><a href="#02-02-2016"><hr><h2><b>February 2, 2016</b></h2></a>
 585<blockquote><p>"I checked it very thoroughly," said the computer, "and that
 586quite definitely is the answer. I think the problem, to be quite honest with
 587you, is that you've never actually known what the question is."
 588- The Hitchhiker's Guide to the Galaxy.</p></blockquote>
 589
 590<p><a href=downloads/toybox-0.7.0.tar.gz>Toybox 0.7.0</a>
 591(<a href=https://github.com/landley/toybox/releases/tag/0.7.0>git commit</a>)
 592is out.</p>
 593
 594<p>The new commands in defconfig are <b>iotop</b>, <b>top</b>, <b>pgrep</b>,
 595and <b>pkill</b>
 596(most replacing corresponding versions from pending). Added grep -ABC,
 597swapon -d (discard), mkswap -L (label) and UUID support, and find -delete.
 598Izabera added free -h and unshare -f. Josh Gao implemented tail -f.
 599Jose Bollo submitted cp --preserve=context,attr. Kylie McClain added
 600mktemp -u.</p>
 601
 602<p>In pending there's the start of a vi command, and Sameer Pradhan contributed
 603a new dhcp6. This cycle saw several rounds of route cleanup and a little dhcp
 604cleanup, but neither are complete yet. Lipi Lee did some cleanup to netstat.c
 605and Elliott Hughes removed warnings from traceroute.</p>
 606
 607<p>Lots of updates to ps: several new -o options, -k (--sort) -O and -M,
 608improved compatibility with Android's historical behavior, and
 609extensive internal code cleanup (including the removal of all
 610the magic constants).</p>
 611
 612<h3><b>Website</b></h3>
 613
 614<p>Dreamhost restored the <a href="#12-21-2015">missing 11 months</a>
 615to the mailing list archive, in the process deleting the month after
 616that. Now they've asked if I have mbox files archiving the new
 617gap (between December 20, 2015 to January 21, 2016, and presumably they
 618could also fill in the gap from December 14, 2014 to January 3, 2015 that's
 619been there since the last time they did this),
 620but due to some gmail filtering I've
 621<a href=http://landley.net/notes-2012.html#15-10-2012>never
 622been able to disable</a>, my copy of those files is spread among 3 different
 623mbox files I'd have to sort/filter/collate. (It's on the todo list.)</p>
 624
 625<p>Added a code of conduct to the README (we're
 626<a href=https://engineering.twitter.com/opensource/code-of-conduct>borrowing twitter's</a>) because somebody
 627made it necessary.</p>
 628
 629<h3><b>Bugfixes</b></h3>
 630<p>Fixed another sed bug where any ] right after [ was skipped (not just the
 631first one in the range, so [[] didn't terminate). Fixed sort -f and added test cases.
 632Assume 80 columns in "ls -m | cat", ls -L is no longer backwards,
 633and ls of files with no paths no longer uses an uninitialized (zero) dirfd.
 634Several bugfixes
 635to find (Gilad Arnold fixed -perm, Daniel K. Levy fixed "find . -exec echo {}",
 636and while we're there I fixed find --prune, made "find . -execdir
 637echo {} + -execdir ls {} +" work, and ripped out the environment size
 638measuring code that checked for a 128k limit removed back in linux 2.6.22).
 639Elliott Hughes fixed the date command's parsing of 4 digit
 640years and documented the %s escape, fixed hwclock -u, and pointed out
 641that runcon needs to exec to do its job (not recursively call another
 642command_main() in the same process). Tom Marshall reported that blkid was
 643handling ext2 wrong. Mike Moreton corrected cpio extraction's uid and gid
 644values, and added a --no-preserve-owner option. Fixed the SUID permission
 645dropping logic (which was a bit over-zealous, preventing some commands from
 646running at all).</p>
 647
 648<p>I'm told that debian-testing broke its libc so the nsenter build breaks,
 649but my attempts to install the debian-testing network cd image under
 650qemu keep breaking. Maybe someday they'll fix it enough I can actually
 651reproduce the problem. (Debootstrap under unbuntu builds an ubuntu-flavored
 652chroot in which toybox builds fine.)</p>
 653
 654<h3><b>Documentation</b></h3>
 655<p>Rewrite of the about.html page, tweaks to design.html, and a re-triage of
 656sbase in roadmap.html. Update to mkstatus.py to collate multiple span
 657tags with the same id, resulting in a larger status.html page (which
 658was previously ignoring some commands in the roadmap).</p>
 659
 660<p>Expanded the defconfig/allyesconfig/allnoconfighelp text in "make help"
 661to explain what they're for.</p>
 662
 663<h3><b>infrastructure</b></h3>
 664<ul>
 665<li><p>Expanded toys.optargs to 64 bits so a command can have more than 32 options.</p></li>
 666<li><p>Added NOEXIT() wrapper to turn xwrap() functions into warning versions
 667using the existing longjump(toys.rebound) infrastructure.</p></li>
 668<li><p>Renamed dirtree->data to dirfd and stopped storing symlink length
 669into it (this fixed a bug where following symlinks to directories
 670didn't give a valid directory filehandle, noticeable with ls -Z).</p></li>
 671<li><p>New TAGGED_ARRAY() infrastructure generates index and bitmask macros
 672for arrays of structures starting with a name string.</p></li>
 673<li><p>New lib/linestack.c for utf8 fontmetrics (draw_str() and utf8len()
 674and so on), and for tracking multiple lines of text
 675(vi, less, shell history) that need wordwrapping and scrolling up/down.</p></li>
 676<li><p>Upgrades to lib/interestingtimes.c: scan_key() now has a timeout
 677in miliseconds and recognizes more sequences including ANSI
 678window size probes. New utf8 test files in tests/files/utf8 including
 679sequence reversing, stacked combining chars, and all three types of
 680unprintable sequences (low ascii <32 ala ^X, invalid utf8 sequences ala
 681<AB><CD>, and invalid unicode code points ala U+1234).</p></li>
 682<li><p>More comma handling code in lib.c: comma_args()</p></li>
 683<li><p>Added error_msg_raw() to shut up fortify's endless static checking false
 684positives.</p></li>
 685<li><p>readfileat() can now realloc() in a loop to read long files
 686("zcat | insmod" needed it).</p></li>
 687</ul>
 688
 689<h3><b>Roadmap</b></h3>
 690<p>We're getting close to having a self-hosting development environment
 691using toybox for the command line. The remaining busybox commands in
 692<a href=http://landley.net/aboriginal/about.html>Aboriginal Linux</a> are:</p>
 693
 694<blockquote><p><b>
 695awk bunzip2 bzcat bzip2 dd diff expr fdisk ftpd ftpget ftpput gunzip gzip
 696less ping route sh sha512sum tar test tr unxz vi wget xzcat zcat
 697</b></p></blockquote>
 698
 699<p>And the remaining non-busybox commands in Aboriginal Linux's build/host
 700directory (from the distcc, genext2fs, e2fsprogs, zlib, and squashfs packagesi)
 701are:</p>
 702
 703<blockquote><p><b>
 704mke2fs fsck.ext2 resize2fs distcc genext2fs unsquashfs distccd mksquashfs tune2fs
 705</b></p></blockquote>
 706
 707<p>Squashfs and distcc are probably out of scope for toybox, but mke2fs,
 708fsck.ext2, resize2fs, genext2fs, and tune2fs should all be added to the
 709above "busybox" replacement list.</p>
 710
 711<p>Remind me to include this countdown in future releases. Once they've all
 712been replaced, the next goal is <a href=http://landley.net/aboriginal/about.html#selfhost>building AOSP under itself</a>.</p>
 713
 714<p>See the full <a href=roadmap.html>roadmap</a> and <a href=status.html>status</a>
 715pages for more details.</p>
 716
 717<a name="12-21-2015" /><a href="#12-21-2015"><hr><h2><b>December 21, 2015</b></h2></a>
 718
 719<p>Yes, 11 months have gone missing from the mailing list web archive.</p>
 720
 721<p>Yesterday evening Dreamhost's mailman server went down (timing out trying
 722to connect). I poked them about it, they
 723<a href=https://twitter.com/landley/status/678781271670149121>blamed
 724DNS</a>, I explained that the hang was _after_ the DNS lookup and
 725entered the dig info into the trouble ticket showing the IPs the DNS
 726queries were returning, they reinstalled the server at that IP from what I
 727assume was their most recent backup, and that's how 11 months of messages
 728vanished out of the archive.</p>
 729
 730<p>I've <a href=https://twitter.com/landley/status/679114451975467008>continued to poke them about it</a> but I honestly believe that's the best they
 731can do. <a href=http://landley.net/dreamhost.txt>Last time</a>
 732this sort of thing <a href=http://landley.net/dreamhost2.txt>happened</a>
 733we went back and forth for months, so I added a link to a backup web
 734archive (in the nav bar on the left) that isn't controlled by dreamhost,
 735and thus doesn't gratuitously lose data on a regular basis. (I note
 736the earlier hole in Dreamhost's archive was never fixed either. That
 737was data never getting archived, this is a year's worth of data that
 738was in the archive until yesterday vanishing after the fact.)</p>
 739
 740<p>If you're wondering why the <a href=http://lists.landley.net>top level</a>
 741list page has been "temporarily disabled" for multiple years now... you'd have
 742to ask Dreamhost. I know I have. More than a dozen times.</p>
 743
 744<a name="03-11-2015" /><a href="#03-11-2015"><hr><h2><b>November 3, 2015</b></h2></a>
 745<blockquote><p>"Alright," said Ford. "How would you react if I said that I'm
 746not from Guildford after all, but from a small planet somewhere in the vicinity
 747of Betelgeuse?" Arthur shrugged in a so-so sort of way. "I don't know," he
 748said, taking a pull of beer. "Why - do you think it's the sort of thing you're
 749likely to say?" - The Hitchhiker's Guide to the Galaxy.</p></blockquote>
 750
 751<p><a href=downloads/toybox-0.6.1.tar.gz>Toybox 0.6.1</a>
 752(<a href=https://github.com/landley/toybox/releases/tag/0.6.1>git commit</a>)
 753is out.</p>
 754
 755<p>We have a new <b>ps</b> command with all the -o fields posix wants (although
 756it doesn't accept BSD non-dash option syntax yet), and <b>bunzip2</b> (not just
 757bzcat but the proper extract-in-place command).
 758Sameer Pradhan added <b>hostid</b> and <b>fsync</b>.
 759Elliott Hughes added <b>flock</b>.
 760
 761<p>The people waiting for <b>human readable number support</b> (du -hH, ls -h,
 762and so on) can thank Elliott Hughes for implementing it. (Our output doesn't
 763exactly match others' because we our "binary" mode will say 1.0G instead of
 7641024M, which is a bug in the other one we didn't emulate.)</p>
 765
 766<p>The other big news is <b>nommu support</b>, tested on the new
 767<a href=http://nommu.org/jcore>jcore</a> processor but presumaby working
 768on any nommu system. A few commands don't support nommu yet, but those
 769are disabled by dependencies on TOYBOX_FORK in menuconfig when building
 770for nommu. The roadmap now has a large section analyzing the uClinux
 771project (note that <a href=http://nommu.org>nommu.org</a> is slowly replacing
 772<a href=http://uclinux.org>uclinux.org</a> as the standard repository of
 773all knowledge and wisdom about nommu. The old site <a href=#12-02-2012>contains
 774much that is apocryphal</a>, or at least wildly inaccurate, and the new one
 775is trying to improve on that).</p>
 776
 777<p>Both "make change" and scripts/single.sh (for building standalone commands
 778without the multiplexer logic) now use the top level .config
 779for toybox global settings such as Linux Security Blanket Module selection,
 780(so make defconfig before change now).</p>
 781
 782<p>Documentation updates to the <a href=code.html>code</a> and
 783<a href=roadmap.html>roadmap</a> pages.</p>
 784
 785<h3>pending</h3>
 786
 787<p>In the pending directory Sameer Pradhan added tftp,
 788and Elliott Hughes sent lsof. Isaac Dunham upgraded mdev,
 789reboot, init, login, and modprobe, and fixed a distro-specific build break in
 790scripts/mkflags.h. Elliott Hughes and Lipi Lee made netstat -p handle
 791command lines longer than 21 characters, and Elliott fixed netstat -e and
 792some build warnings. Yeongdeok Suh fixed a warning in dhcpd.
 793I started cleanup on pgrep/pkill.</p>
 794
 795<h3>Command updates, bugfixes, and infrastructure</h3>
 796
 797<p>The multiplexer's "command not found" error exit is now 127, so now you can't
 798distinguish between a command not being found in the multiplexer and
 799the multiplexer itself not being found by the shell, because people wanted
 800that for some reason.</p>
 801
 802<p>Elliott Hughes made date reject invalid dates rather
 803than set the clock to something weird (setting the clock 100 years into the
 804future makes most Linux desktops surprisingly unhappy, and ntpdate won't fix it
 805either), fixed several ls -l display issues (user/group field ordering,
 806make user/group/lsmcontext left aligned), did the aforementioned
 807extensive work on human readable number output, fixed ionice's default
 808class, fixed a mv overwrite bug, made df's columns auto-size, added
 809--ppid and -Z to ps, and teamed up with Daniel K. Levy to fix
 810a segfault in find's handling of -newer -group or -user.</p>
 811
 812<p>Hyejin Kim added stat -c %T support. Colin Cross worked
 813on vmstat fixing
 814a header printing bug and calculating the bi and bo columns in the right
 815units. Isabella Parakiss reported that sed -e "/x/c\" -e "y" added an extra
 816newline and that grep -w '\(x\)\1' didn't work, both now fixed.
 817Alistair Strachan fixed several problems with switch_root. Kylie McClain
 818pointed out env should be able to clear variables via NAME= syntax.
 819Dima Krasner added support for running blkid without a partition (so it shows
 820all partitions). Hyejin Kim sent in a bunch of static analysis bug reports.</p>
 821
 822<p>Isabella Parakiss reported that sed -e "/x/c\" -e "y" added an extra
 823newline and that grep -w '\(x\)\1' didn't work, both now fixed.
 824Alistair Strachan fixed several problems with switch_root. Kylie McClain
 825pointed out env should be able to clear variables via NAME= syntax.
 826Dima Krasner added support for running blkid without a partition (so it shows
 827all partitions). Hyejin Kim sent in a bunch of static analysis bug reports.</p>
 828
 829<p>Two large thinko fixes in oneit: -3 was always enabled (which would
 830eventually block if the child never read the exiting PID numbers from its file
 831descriptor #3 until the pipe filled up), and the signal handlers weren't
 832set up right (for requesting semi-graceful halt/poweroff/reboot).
 833Calling install without a mode is now 0755, and install -g 0 no longer clashes
 834with cp --preserve. Better error message for ls -r on unreadable
 835directories, and ls -Z now uses O_PATH (with the /proc/self/fd/%d
 836workaround for kernel stupidity as necessary).</p>
 837
 838<p>Date now understands @unixtime[.fraction] and uses -D for
 839the set-side format (matching busybox's extension for this). The seq -f
 840string now checks that it's got exactly one %f escape with the correct
 841attributes (and a whole bunch of test cases for it). Fixed a bug
 842in od that screwed up the position indicator on arm and mips.
 843In stat the d/h units moved from %d %D to the default string.
 844And patch can now correctly apply hunks with trailing context to the start of
 845the file.</p>
 846
 847<p>The prompt argument moved out of yesno() (the caller can print the prompt
 848themselves). Replaced toys.exithelp with help_exit(). Added new
 849XVFORK() macro, and xpopen_both() calls /proc/self/exe when passed
 850a NULL argv (see cpio -p for example usage). Replaced toys.recurse
 851with toys.stacktop so the recurse or re-exec decision is now based
 852on bytes of stack space used. Marked a bunch of command-local functions
 853static.</p>
 854
 855<p>New additions to lib/ include strlower(), xconnect(), and the
 856aforementioned help_exit().
 857The testsuite now has some infrastructure tests based on "example"
 858commands such as toys/examples/test_human_readable.c.
 859The login command finally got a long-overdue cleanup (it's one of the
 860commands that predate the "pending" directory but were part of the reason
 861for it). Hexedit had an
 862uninitialized variable (of course gcc didn't spot it, it was too busy
 863warning about "may be used uninitialized but never actually is" variables).</p>
 864
 865<p>Tweaked makefile so
 866"make CROSS_COMPILE=prefix-" (as well as "CROSS_COMPILE=prefix- make",
 867which still works). Toybox is now installed chmod -w so broken installers
 868(like the bunzip2 package's) that try to overwrite existing binaries won't
 869knock out the whole of toybox.
 870GCC 5.2.0 stopped being able to compile Linux 2.6.12's kconfig, but
 871we added a workaround. You can now build uptime without utmpx.h.
 872Alejandro Joya pointed out that enabling smack required smack on the host
 873as well as target when cross compiling, which is now fixed.</p>
 874
 875<p>Note: toybox can autodetect nommu support when building with a uClibc
 876toolchain such as <a href=http://landley.net/aboriginal/downloads/binaries/old/1.4.3/cross-compiler-sh2eb.tar.gz>the one from Aboriginal Linux</a>,
 877but <a href=http://github.com/richfelker/musl-cross-make>with musl-libc</a>
 878you'll have to enable CONFIG_TOYBOX_MUSL_NOMMU_IS_BROKEN to work around the
 879fact they provide a non-functional fork() implementation that always returns
 880-ENOSYS, to prevent you from compile-time probing for nommu support when
 881cross-compiling. Unfortunately "preventing you from probing" seems to be
 882an explicit policy with musl, they also don't provide an "#ifdef __MUSL__"
 883because their library is perfect and you're only ever allowed to work around
 884other people's bugs, not theirs. So we have to use menuconfig to manually
 885enable musl-specific bug workarounds.</p>
 886
 887<a name="23-07-2015" /><a href="#23-07-2015"><hr><h2><b>July 23, 2015</b></h2></a>
 888<p>I recreated the <a href=downloads/toybox-0.6.0.tar.gz>0.6.0 source tarball</a>
 889(new sha1sum 08fb1c23f520c25a15f262a8a95ea5b676a98d54)
 890because I forgot to add --prefix to the git archive command when I updated
 891my release script from mercurial, so the files weren't in an enclosing
 892directory. (Ooops.)</p>
 893
 894<a name="19-07-2015" /><a href="#19-07-2015"><hr><h2><b>July 19, 2015</b></h2></a>
 895<blockquote><p>
 896The reason why it was published in the form of a micro sub meson electronic
 897component is that if it were printed in normal book form, an interstellar
 898hitchhiker would require several inconveniently large buildings to carry it
 899around in." - The Hitchhiker's Guide to the Galaxy </p></blockquote>
 900
 901<p><a href=downloads/toybox-0.6.0.tar.gz>Toybox 0.6.0</a>
 902(<a href=https://github.com/landley/toybox/releases/tag/0.6.0>git commit</a>)
 903is out. (Yes, git. See the <a href=#05-04-2015>previous news entry</a>.)</p>
 904
 905<p>Sorry for the unusually long gap between releases. Since last release Ye
 906Olde Project Maintainer traveled to japan twice and had two more "once
 907a century" floods at home. (Probably a coincidence.) Still catching up.</p>
 908
 909<h3><b>CELF/ELC talk and Wikipedia[citation needed] article</b></h3>
 910
 911<p>I gave another State Of The Toybox talk
 912(<a href=https://www.youtube.com/watch?v=04XwAbtPmAg>video</a>
 913<a href=http://landley.net/talks/celf-2015.txt>outline</a>), in which I
 914repeat my <a href=http://landley.net/notes-2013.html#07-11-2013>perennial</a>
 915<a href=https://twitter.com/landley/status/557309224535851009>complaint</a>
 916that Wikipedia[citation needed]
 917<a href=http://en.wikipedia.org/wiki/Toybox>still</a>
 918<a href=https://en.wikipedia.org/wiki/BusyBox#Controversy_over_Toybox>says</a>
 919toybox was relicensed before its hiatus, when relicensing was why
 920the hiatus ended.</p>
 921
 922<p>Since Wikipedia[citation needed] seems unable to do the
 923<a href=#15-11-2011>most</a>
 924<a href=http://landley.net/hg/toybox/log/tip/LICENSE>basic</a>
 925<a href=http://landley.net/notes-2011.html#13-11-2011>research</a> on
 926this point, and has stuck to an incorrect sequence of events for years,
 927I've been gradually escalating my attempts to correct them. Toybox
 928came out of mothballs in November 2011 <b>because</b> it could be
 929relicensed. That's what opened up a new niche busybox wasn't already
 930filling with a 10 year headstart.</p>
 931
 932<a name="asterisk_back" />
 933<p>The article has plenty of smaller issues<a href=#asterisk>*</a>, but
 934given that I gave an entire talk at Ohio LinuxFest in 2013
 935(<a href=http://landley.net/talks/ohio-2013.txt>outline</a>,
 936<a href=https://archive.org/download/OhioLinuxfest2013/24-Rob_Landley-The_Rise_and_Fall_of_Copyleft.mp3>audio</a>) on why I switched away from GPL for
 937my projects, that one bugs me.</p>
 938
 939<h3><b>New stuff this release</b></h3>
 940
 941<p>There's a new android menu in menuconfig, and rather a lot of Linux
 942Security Module support (Smack for Tizen from Xavier Roche and José Bollo,
 943and SELinux for Android from Elliott Hughes; see
 944the Security Blanket menu under global settings in menuconfig) has
 945trickled in, although there's still more to come.</p>
 946
 947<p><b>New commands:</b> Added reset, nproc, ionice, and iorenice.
 948Elliott Hughes contributed xxd, runcon,
 949restorecon, load_policy, getenforce, setenforce, getprop, and setprop.
 950Promoted shred, nsenter, and hwclock.</p>
 951
 952<p>You can once again build catv now the flag infrastructure's been updated to
 953let it coexist with cat -v.
 954And on a long plane flight I wrote
 955hexedit, an interactive hex editor that implements the start of
 956cursor control infrastructure (for eventual use by less and vi and shell
 957command history and so on).</p>
 958
 959<p><b>New options:</b> Added sed -E as a BSD-compatible synonym for -r.
 960Upgraded oneit with -r (restart), -3 (send exiting PID values to child),
 961and signal handling. Added -v option to timeout, -m to mknod, -u to shred,
 962-t to dmesg, and -123 to head and tail. Added implicit "." to grep -r without
 963any files to work on. Hyejin Kim requested prefix support for truncate -s.
 964Greg Hackman added -inum to find.
 965Jan Cybulski added the smack side of ls -Z support. Various patches also
 966added -Z to mkdir, mknod, and mkfifo.
 967Basic cp --preserve support went in, but not yet the xattr/LSM parts.</p>
 968
 969<p>The toybox command now has a --version option,
 970which uses "git describe" if available.</p>
 971
 972<p><b>Build infrastructure:</b>
 973The "make change" target now saves the output of each failed standalone
 974command build in a .bad file, and "make defconfig" is quieter now.</p>
 975
 976<p>Paul Barker submitted a large patch changing command install paths so
 977"toybox can be installed alongside busybox without confusing
 978update-alternatives". (There's some argument over
 979what the right paths should be, and I'm waiting for
 980people to tell me what else needs fixing because I have no idea. I've
 981been symlinking /bin to /usr/bin since 2002
 982<a href=http://landley.net/writing/hackermonthly-issue022-pg33.pdf>for
 983historical reasons</a>.)</p>
 984
 985<p><b>Docs:</b> The repository link now goes to github, with another link
 986to the commit rss feed.</p>
 987
 988<p>Elliott Hughes updated the Android section of the roadmap
 989(and he would know). Redid bits of scripts/mkstatus.py to make updating
 990status.html easier, and the README is larger.</p>
 991
 992<p>More description of option parsing in code.html, which now describes the
 993FLAG_x macros, switching flag macro sets with FOR_newcommand, how
 994configuration zeroes flag macros and using FORCE_FLAGS to suppress the
 995zeroing of options shared between commands. Also added description of ";"
 996to make --longopts take an optional =value part, and more about TOYBOX_DEBUG
 997to check NEWTOY() option strings (otherwise a bad option string makes
 998lib/args.c obviously segfault, but doesn't explain why).</p>
 999
1000<p>Added a "Why 0BSD?" section to license.html when submitting zero clause bsd
1001to SPDX (according to the pending license spreadsheet, it's been approved for
1002SPDX 2.2).</p>
1003
1004<p>The old list of commands needing cleanup but not in pending was
1005removed from toys/pending/README and instead the issues were added
1006as TODO comments in the individual commands.</p>
1007
1008<p><b>Bugfixes:</b>
1009Fixed mount -a segfaulting without -O (reported by Janus Troelsen),
1010and made it try a "become rw" ioctl() on the block device before falling
1011back to mounting read only (because Android expects that).
1012Fixed printf -- and printf ---. Lots of tweaks to ls -l spacing with
1013different options. Make touch -d and -t actually set time when you don't
1014specify nanoseconds.
1015Fixed a subtle bug where recursive calls (toybox commands that run other
1016toybox commands) weren't resetting all their state. (This manifested as
1017a "no }" error from "find | xargs sed", but could cause other problems.)
1018And David Halls reported another sed bug trying to compile libiconv (which
1019left extra \ at the start of lines in a generated shell script, breaking
1020the build). Output an error message for "cat /mnt".</p>
1021
1022<p>Kylie McClain reported that mktemp broke when $TMPDIR was set to an empty
1023string (which is not the same as unset), that install/find didn't support
1024numeric uid/gids, and that sort -z affects both input and output.
1025Isabella Parakiss fixed a printf.c bug.
1026David Halls fixed bugs in install -D and find -exec. Samuel Holland
1027fixed unshare -r. Hyejin Kim fixed makedevs with a count of 1, fold -w
1028range checking, an error path in scripts/mkflags.c, added -i to dhcpd,
1029and stopped su from prompting the root user for the new user's password.
1030Jan Cybulski spotted wrong indentation when combining ls -s and -i with -C and
1031-x. José Bollo fixed stat %G. Sameer Pradhan fixed a bug in mkfifo -Z.</p>
1032
1033<p>Elliott Hughes asked for a default SIGPIPE handler to disable
1034the signal handler bionic's dynamic loader installs (yes really). Still not
1035100% sure what the correct behavior is there. (Posix is
1036(<a href=http://permalink.gmane.org/gmane.comp.standards.posix.austin.general/10915>actively unhelpful</a>, but at least they're taking
1037<a href=http://austingroupbugs.net/view.php?id=789#c1976>years to
1038make up their mind</a>. Elliott also sent patches to fix a typo in
1039useradd.test, add missing arguments to error_exit() calls and clean up
1040printf() format strings, fix an off by one error in human_readable(),
1041fix dmesg -c error reporting, fix a segfault in comma_scan where the option
1042was the last item in optlist (triggered by mount -o ro,remount), fix
1043hwclock -w, made ifconfig print lowercase MAC addresses (it was bothering
1044him), and make terminal_size() read the right environment variable
1045(LINES, not ROWS). And he suggested the test suite notice high command exit
1046values (corresponding to segfault or other signals).</p>
1047
1048<p>People are apparently using toys/pending commands, despite the police tape
1049and flashing lights, so added louder warnings to toys/pending/README.
1050Elliott Hughes fixed various problems with tar, dd, more, and top.
1051Hyejin Kim cleaned up syslogd and dumpleases. Isaac Dunham added hotplug
1052support to mdev. Yeongdeok Suh added RFC-3315 ipv6 support to dhcpd.</p>
1053
1054<p>I rewrote ps.c from scratch (in pending), but it's not ready for real use
1055yet.</p>
1056
1057<p><b>Portability:</b>
1058On the portability front Bernhard Rosenkranzer fixed a problem where the
1059menuconfig code wouldn't compile in C99 mode. (This led to me documenting
1060the craptacular nature of kconfig in a README, and the plan to replace it
1061sometime before 1.0.) Some extra flags to shut up overzealous llvm warnings
1062were added (and have to be probed for because gcc complains about
1063arguments it doesn't recognize even when they switch stuff _off_ using
1064a standard syntax). Don't depend on malloc(0) to return non-null in ls.
1065David Halls fixed some mac/ios portability issues,
1066implying somebody's built at least part of toybox on a mac.</p>
1067
1068<p>Added basename_r() to lib/lib.c because the posix semantics for basename()
1069are stupid but what the gnu guys did to it was appalling.
1070Turns out bionic already had a basename_r(), but posix still doesn't.
1071Fixed it up in portability.h, but this
1072could break more stuff in future. (Correct fix is to lobby posix to add it,
1073which would probably take about 15 years...)</p>
1074
1075<p><b>Infrastructure:</b>
1076The build now checks $LDFLAGS for linker-only flags, and allows the strip
1077command to fail (binflt toolchains provide a strip that doesn't work).
1078Since time.c uses floating point, added TOYBOX_FLOAT dependency in config.</p>
1079
1080<p>There's a lib/lsm.h defining varous inline functions for linux
1081security modules stuff, if (lsm_enabled()) should turn into a compile-time
1082constant 0 and let code drop out when TOYBOX_LSM_NONE selected, but
1083testing against CFG_TOYBOX_LSM_NONE or derived symbols is still useful
1084becuase when it _is_ enabled the probe turns into a system call you
1085don't want to repeat too much.</p>
1086
1087<p>Switched a bunch of commands from signal() to xsignal(). Factored out
1088xgetgrnamid() and xgetpwnamid() into xwrap.c. Make time.c depend on
1089TOYBOX_FLOAT (since it always uses float so shouldn't be available on
1090build targets without even software float). Added readfileat() to lib/lib.c.</p>
1091
1092<p>The dirtree infrastructure now passes in full flags for the old symlink
1093field, and the new DIRTREE_SHUTUP flag disables warnings if a file vanishes
1094out from under you during traverse. New dirtree_start() wrapper to
1095create dirtree root with only two arguments.</p>
1096
1097<p>The not-curses infrastructure introduced by hexedit mostly moved to
1098lib/interestingtimes.c.</p>
1099
1100<a name="asterisk" />
1101<a href="#asterisk_back" />Asterisk:</a> such when
1102Tim contacted me (my blog says a couple days before nov 13, 2011, I.E.
110311/11/11 not some specific day 2 months later) to ask if I wanted to work
1104on a new project he was proposing called
1105<a href=http://www.elinux.org/Busybox_replacement_project>BentoBox</a>
1106(because I used to do busybox, he'd forgotten toybox existed
1107until I brought it up). And don't ask me what "focuses not on compatibility
1108with its GNU counterparts" means when CP_MORE adds 7 non-posix options
1109and toys/other has 84 commands in neither posix nor LSB. I think they're
1110struggling to explain the difference having dismissed "licensing" as being
1111the reason it started up again after a long hiatus? The reason I don't think
1112GNU is special is there are a half-dozen other independent
1113implementations of the same unix command tools out there (AT&amp;T,
1114BSD, Coherent, Minix, plan 9, busybox, toybox, and several more analyzed in
1115the <a href=roadmap.html>roadmap</a>, and that's ignoring the implementations
1116written for DOS or in assembly over the years). But I do care what
1117Linux From Scratch expects, and if it's
1118<a href=http://archive.linuxfromscratch.org/lfs-museum/7.6/LFS-BOOK-7.6-NOCHUNKS.html#ch-tools-gcc-pass1>calling mv -v</a>
1119then I impelement mv -v
1120even if <a href=http://landley.net/toybox/roadmap.html>posix hasn't got
1121it</a>. And I don't know why "gnu counterparts" would describe this when
1122util-linux isn't a gnu package, nor are info-zip, e2fsprogs, kmod, less,
1123procps, shadow, sysklogd, vim, zlib, sudo, dhcpcd...</p>
1124
1125<a name="05-04-2015" /><a href="#05-04-2015"><hr><h2><b>April 5, 2015</b></h2></a>
1126<p>Since <a href=https://android.googlesource.com/platform/external/toybox/>android</a> and
1127<a href=https://git.tizen.org/cgit/platform/upstream/toybox.git>tizen</a>
1128and <a href=https://github.com/kraj/meta-musl/tree/master/recipes-core/toybox>openembedded</a>
1129and <a href=https://packages.gentoo.org/package/sys-apps/toybox>gentoo</a>
1130and so on have all been using Georgi Chorbadzhiyski's git mirror rather
1131than the mercurial repository, I bit the bullet and switched the project's repo
1132<a href=https://github.com/landley/toybox>to git</a>. Georgi's
1133<a href=https://github.com/gfto/toybox>mirror</a> is now pulling from that.</p>
1134
1135<a name="25-02-2015" /><a href="#25-02-2015"><hr><h2><b>February 25, 2015</b></h2></a>
1136<blockquote><p>"A common mistake that people make when trying to design
1137something completely foolproof is to underestimate the ingenuity of
1138complete fools."</p><p>- The Hitchhiker's Guide to the Galaxy.</p></blockquote>
1139
1140<p><a href=downloads/toybox-0.5.2.tar.gz>Toybox 0.5.2</a>
1141(<a href=/hg/toybox/shortlog/1702>commit 1702</a>) is out.</p>
1142
1143<p>New promoted commands: sed (finally fixed enough it builds Linux From
1144Scratch), printf (cleaned up and promoted), shred and
1145base64 (the Tizen guys wanted them), getenforce, setenforce, and chcon (android),
1146mix (promoted with fixes from Isaac Dunham), nsenter (from
1147Andy Lutomirski, merged into unshare).</p>
1148
1149<p>Elliott Hughes submited a bunch of patches to support Android (to
1150both toybox and Bionic libc, which he maintains). On toybox's end this
1151involved a lot of fixups to portability.[ch] and fixes to over a dozen
1152commands, plus several new ones. Other portability fixes included working
1153with buildroot's uclibc fork and building for nommu targets.</p>
1154
1155<p>The new "make change" target builds each toybox command as a standalone
1156binary. Rather a lot of commands that didn't build by themselves (mv depending
1157on cp and so on) were hit with a large rock until they built standalone.
1158This involved rewriting bits of option parsing, more elaborate dependency
1159generation, making each command have its own config
1160symbol and main() function (even when it's just a wrapper calling another
1161command's main()), and so on. Also, some commands can't be built standalone
1162at a conceptual level: "help" describes other enabled commands and "sh"
1163has a number of bulitin commands (cd, exit, set) that require the
1164multiplexer infrastructure, so "make change" filters them out.</p>
1165
1166<p>The mailing list's web archive is still screwed up. Dreamhost has
1167been trying to fix it since approximately September. There are
1168<a href=http://www.mail-archive.com/toybox@lists.landley.net/>two</a>
1169<a href=http://news.gmane.org/gmane.linux.toybox>other</a> less broken
1170archives, but neither has quite the same UI as mailman.</p>
1171
1172<h3>Bugfixes and tweaks</h3>
1173
1174<p>Cynt Rynt sent in tests for ifconfig,
1175Robert Thompson taught factor to accept whitespace separated arguments,
1176Hyejin Kim pointed out that some of mktemp's longopts were attached to
1177the wrong short options,
1178Luis Felipe Strano Moraes fixed a wrong free() call in bootchartd in pending.
1179Patches from Ashwini Sharma to make "df /dev/node" work, prevent du from
1180looping endlessly following symlinks, and to make expr.c
1181(in pending) understand == and regex matches. (Speaking of expr, it gets
1182priority groupings wrong but the bug was actually in the posix spec's
1183HTML conversion. They fixed the posix spec upstream for us. Still need
1184to fix the expr code, but it's in pending for a reason...)</p>
1185
1186<p>Some commands grew new option flags, such as cp --remove-destination
1187and touch -h.</p>
1188
1189<p>The parallel build has better error reporting now. When toybox needs to
1190re-exec itself to regain suid root permissions and hasn't got the suid bit,
1191it now gives the right error message ("not root" instead of "no such command").
1192
1193<p>Added a test to "mount" to not mount the same device/directory combination
1194over itself (the OS catches this for block devices, but not for tmpfs).
1195Make blkid distinguish ext3 from ext4. Added catv back into cat (because
1196the Android guys wanted it, and they have historical usage on their side,
1197so...). Handle nanoseconds in touch.</p>
1198
1199<p>Fixed a segfault when CP_MORE was disabled (the resulting option flag list
1200no longer defined -d but still had it in option groups at the end).
1201Workaround for glibc redefining dirname() and basename() to random non-posix
1202semantics because gnu. (They could have created dirname_r() but didn't want
1203to.)</p>
1204
1205<p>Fix an ifconfig test that was preventing assigning an ipv4 address to
1206interface aliases. Several cleanup passes on hwclock but not quite
1207promoted out of pending yet.<p>
1208
1209<p>Fixed a wrong error message in rm (if you had a chmod 000 directory and
1210did rm -r on it without -f, after the prompt it would complain it was a
1211directory, which was not the problem).</p>
1212
1213<p>The gzip compression code now does "store only" output to stdout, for
1214what that's worth.</p>
1215
1216<p>Cleanup mountpoint and expand, and remove them from toys/pending/README
1217(a list of commands that predate the toys/pending directory but needed
1218another pass).</p>
1219
1220<h3>Library and infrastructure:</h3>
1221
1222<p>Reworked the option parsing infrastructure so more commands build
1223standalone (via scripts/single.sh or "make change"). The option flag bit
1224values are no longer packed, it leaves spaces where currently disabled
1225flags go, and you can #define FORCE_FLAGS so disabled flags aren't zeroed.
1226This allows multiple commands to more easily share infrastructure, even if
1227your current flag context is for a disabled command (switched off in config),
1228you can force them to stay on and as long as the flags read the same right
1229to left they'll have the same values.</p>
1230
1231<p>We've started removing use of strncpy() because it's a hugely broken
1232standard C function: the length is the maximum length to _append_, not
1233the size of the destination buffer. It memsets the remaining space it didn't
1234copy ala "memset(dest+strlen(dest), 0, len);" so
1235if you think len is the size of dest you're guaranteed to stomp memory off the
1236end). And if it runs out of space it won't null terminate because reasons.
1237(Meanwhile sprintf("%*s", len, str) is counting wide characters in your current
1238locale, so if you set a locale other than "C" it will also go past your
1239allocated buffer size. Whoever is maintining the C library standards is really
1240bad at strings.)
1241Instead we have xstrncat() which will error_exit() if src+dest+1 doesn't
1242fit in the buffer. (Because randomly truncating input data isn't necessarily
1243an improvement.) And there's always xmprintf().</p>
1244
1245<p>Similarly, strtol() doesn't return an error indicator on overflow,
1246you have to clear and then check errno. So new xstrtol() that cares
1247about overflow.</p>
1248
1249<p>The bionic and musl guys agree faccessat(AT_SYMLINK_NOFOLLOW) is not
1250supported, so stop using it.</p>
1251
1252<p>Fixed toy_exec() to detect when argc is in optargs, so we don't
1253need a separate xexec_optargs().</p>
1254
1255<a name="18-02-2015" /><a href="#18-02-2015"><hr><h2><b>February 18, 2015</b></h2></a>
1256<p>Dreamhost continues to be unable to make mailing list archives work, so
1257here's <a href=http://www.mail-archive.com/toybox@lists.landley.net/>another
1258list archive</a> with a less awkward interface than gmane.</p>
1259
1260<p>(Neither gives you the convenient historical monthly views of mailman,
1261but I still have hopes dreamhost will someday figure out what they're doing
1262wrong. They've only been trying since October. Last month they did a
1263<a href=http://www.dreamhoststatus.com/2015/01/14/discussion-list-hardware-maintenance/>hardware upgrade to fix a software problem</a>, and the stale
1264data loads much faster now, so that's something.)</p>
1265
1266<p>Update (Feb 19): the archive started updating again, by discarding
1267all the pending data. So there are now _two_ giant holes in Dreamhost's
1268web archive, from Dec 15-Jan 3, and then another hole from Jan 16-Feb 18.
1269The relevant messages are in both of the other archives. Here's hoping
1270the chronic archive constipation problem won't happen a sixth time.</p>
1271
1272<a name="30-12-2014" /><a href="#30-12-2014"><hr><h2><b>December 30, 2014</b></h2></a>
1273<p>Due to Dreamhost's <a href=http://landley.net/dreamhost.txt>ongoing</a>
1274<a href=http://landley.net/dreamhost2.txt>inability</a> to make mailman
1275work reliably, I've added a link to a backup web archive at
1276<a href=http://news.gmane.org/gmane.linux.toybox>gmane</a> to the nav bar
1277on the left.</p>
1278
1279<p>You still subscribe to the list through
1280<a href=http://lists.landley.net/listinfo.cgi/toybox-landley.net>the first link</a>.</p>
1281
1282<p>Update (January 27, 2015): they're <a href=https://twitter.com/landley/status/558428839462703104>still working on it</a>.</p>
1283
1284<a name="19-11-2014" /><a href="#19-11-2014"><hr><h2><b>November 19, 2014</b></h2></a>
1285
1286<blockquote><p>"This time it was right, it would work, and no one would have to get nailed to anything." - The Hitchhiker's Guide to the Galaxy.</p></blockquote>
1287
1288<p><a href=downloads/toybox-0.5.1.tar.bz2>Toybox 0.5.1</a>
1289(<a href=/hg/toybox/shortlog/1566>commit 1566</a>) is out.</p>
1290
1291<p>It's an interim release, mostly bugfixes. There are several new commands,
1292but they're all in pending.</p>
1293
1294<h3>Development</h3>
1295
1296<p>Finally implemented sed, which is still in pending because although
1297it's feature complete according to posix, and even passes the parts of
1298Busybox's sed test suite that aren't explicitly testing for gnu bugs we
1299don't want to copy, it's not yet good enough to build Linux From Scratch.
1300(The ./configure stages use very long sed scripts. 20 commits worth of
1301implementation and debugging, just under 1000 lines of code, and there's
1302still more to do. We're definitely up to some of the "fiddly" commands now.
1303Did you know "echo hello | sed p - -" segfaults gnu sed in Ubuntu 12.04?
1304Yeah...)</p>
1305
1306<p>Talked with the Tizen developers to follow up on their desire to
1307make toybox a part of the base Tizen system, and got a list of commands
1308to add to the roadmap. The tizen todo list is:</p>
1309
1310<blockquote><p>
1311wget, sha256*, gzip, gunzip, bunzip2, rsync, zdiff*,
1312less, ar, arch, base64, csplit, dir, fmt, join, 
1313nproc, shred, shuf, stdbuf, stty, test, tr, unexpand,
1314users, vdir, diff3, sdiff, dosfsck (fsck.vfat), awk, fdisk
1315</p></blockquote>
1316
1317<p>(Most of which was already on the todo list, but it helps prioritize.)</p>
1318
1319<p>Fixed md5sum and sha1sum on big endian systems (reported by James McMechan).
1320Andy Lutomirski fixed unshare's help text and option parsing,
1321and submitted nsenter (a tool to use setns(2)) to pending.
1322Isaac Dunham implemented acpi -ctV options, and spotted the bug that ls -d
1323was inappropraitely following command line symlinks without -H or -L (it
1324should act like ls -l does), and ls -F handles symlinks wrong too.
1325Lukasz Szpakowski sent in two bugfixes to tail.c. Cynt Rynt spotted an
1326unnecessary assignment in lib/password.c.</p>
1327
1328<p>Ashwini Sharma's team was as busy as usual, submitting tr, crontab, and
1329ipcrm, and hwclock to pending, more features to the pending ip.c, and a
1330pile of bugfixes (to chgrp, killall, ifconfig, insmod,
1331losetup, comm, cp, id, xwrap, netcat, modprobe, nohup...) mostly found by
1332static analysis. (These fixes are mostly to seldom-used codepaths like the
1333TOYBOX_FREE config option, but test coverage is always appreciated.) Ashwini
1334also suggested upgrading ln -f to leave the original target alone if link
1335creation fails, and reported that mv -f and -i weren't implemented (now fixed).</p>
1336
1337<p>New config option: TOYBOX_NORECURSE prevents xexec() from making internal
1338function calls (for nommu systems with a finite stack).</p>
1339
1340<p>The "toybox" multiplexer command no longer adds a trailing space to each
1341line of command names, so things like "./toybox | tr ' \n' '|'" to create
1342a grep pattern snippet are easier to do. (Why you'd want to is your business,
1343but the output is tidier now.)</p>
1344
1345<h3>Infrastructure</h3>
1346
1347<p>Isaac Dunham added Android support to portability.h, including compile
1348probes for functions missing from bionic-libc, and annotated the commands that
1349use those functions. We haven't really tested building against bionic,
1350but in theory it's possible now.</p>
1351
1352<p>Running the test suite now color codes the PASS/SKIP/FAIL notifications
1353if output is to a tty. (And in case you missed it last time, VERBOSE=fail
1354to stop at the first failure is really useful.)</p>
1355
1356<p>In loopfiles_rw() use O_CLOEXEC instead of O_RDONLY to request the loop
1357function close filehandles for us. (Otherwise the callback function must
1358close each supplied filehandle itself.)</p>
1359
1360<p>The printf-style escape parsing ("\n" and friends) got factored out into
1361a new unescape() function.</p>
1362
1363<a name="02-10-2014" /><a href="#02-10-2014"><hr><h2><b>October 2, 2014</b></h2></a>
1364<blockquote><p>"There is an art, it says, or rather, a knack to flying.
1365The knack lies in learning how to throw yourself at the ground and miss...
1366Clearly, it is this second part, the missing, which presents the
1367difficulties." - The Hitchhiker's Guide to the Galaxy.<p></blockquote>
1368
1369<p><a href=downloads/toybox-0.5.0.tar.bz2>Toybox 0.5.0</a>
1370(<a href=/hg/toybox/shortlog/1512>commit 1512</a>) is out.</p>
1371
1372<h3>New commands</h3>
1373
1374<p>The new commands are find, install, factor, and mount. Promoted commands
1375(cleaned up and moved out of "pending") are lspci, inotifyd, and blockdev.</p>
1376
1377<p>cp now implements -HL and -F to force delete of pending files, cpio now
1378ignores -m and implements -p, ls -C now has utf8 support (using wcwidth
1379instead of strlen), and umount got a number of upgrades involving
1380looking things up in /proc/mounts. Other minor cleanups happend to
1381cut, touch, free, and id.</p>
1382
1383<p>In pending: Bradley Controy submitted mix (adjusts OSS sound volume). Ashwini
1384Sharma submitted diff, userdel, blockdev, ipcs, and crond, upgraded
1385fdisk, fsck, and ftpget, and ran a static analyzer on a lot of other code.
1386Partial cleanup was done to useradd, userdel, groupadd, and groupdel.</p>
1387
1388<h3>Build infrastructure</h3>
1389
1390<p><b>Parallel builds</b></p>
1391
1392<p>The build now takes advantage of SMP, autodetecting the number of
1393processors. (Export the environment variable CPUS to pick a specific number.)
1394Other build changes: split out $LDOPTIMIZE because old compilers complain
1395about linker options passed with -c, and the entire "generated" directory now
1396gets deleted by clean (the README that was in there got merged into code.html).</p>
1397
1398<p><b>Standalone builds</b></p>
1399
1400<p>The standalone build infrastructure (scripts/single.sh) got upgraded to
1401build more commands as standalone executables. In make.sh the source file
1402selection uses a regex to find the source files with the NEWTOY/OLDTOY macro
1403for the command. It enables each command's
1404sub-options (so CP has CP_MORE), enables I18N and FLOAT support to build
1405full-featured commands, and includes --help text (at least when
1406the command doesn't use another command's help). The OLDTOY() macro
1407now produces (redundant) function prototypes so you can build an OLDTOY
1408without the NEWTOY</p>
1409
1410<p>It doesn't quite have complete coverage yet, the defconfig entries that
1411aren't building standalone yet are:</p>
1412
1413<blockquote><p>chown, egrep, fgrep, fstype, halt, mv, nc, poweroff, unix2dos,
1414whoami</p></blockquote>
1415
1416<p>The main reason for standalone build failures is NEWTOY() or OLDTOY()
1417entries that don't have their own config symbol. Another problem is entries
1418that depend on another entry in kconfig, usually because common infrastructure
1419is using one command's flags (which the other commands copy): if that command
1420is disabled, the FLAG macros become 0 so dead code elimination can remove the
1421code. It's <a href=http://landley.net/hg/toybox/rev/1503>possible
1422to untangle</a> this, but a bit awkward. (It boils down to conflicting
1423design goals in the two contexts.)</p>
1424
1425<p>Standalone builds are used by the test suite when testing individual
1426commands.
1427
1428<p><b>Snapshot builds</b></p>
1429
1430<p>A new addition to the "generated" directory is generated/build.sh
1431containing a single compiler command line to build toybox in its current
1432configuration. Combined with the generated/*.{h,sh} files from an
1433exisiting build, this may let you build on a new system that hasn't quite
1434got enough OS bits working to run a full configureand make.</p>
1435
1436<h3>Internals</h3>
1437
1438<p>Library code: xcreate/xopen now O_CLOEXEC by default to avoid leaking
1439filehandles to child processes. DIRTREE_COMEAGAIN's second callback is now
1440done with the directory filehandle still open (new dir->again variable added
1441to distinguish first from second callback, and requesting DIRTREE_RECURSE now
1442requires passing in the specific macro value, not just a true/false).
1443Use daemon() out of libc instead of hand-rolled daemonize() in various
1444pending commands. string_to_mode() now passes through type bits so you can
1445use it to more easily modify a file's existing mode.
1446Split xpopen() into xpopen_both(), xopen(), and xrun() depending on whether
1447we want to redirect both, one, or neither of stdin/stdout.</p>
1448
1449<p>Bugfixes: Better error message when TOYBOX_SUID option can't drop priviliges
1450(which happens when you suid something _other_ than root).
1451The old pending version of nbd_client.c wasn't deleted when the
1452command was promoted (and the build would break if both were enabled),
1453toy_exec() sometimes needs to re-exec from $PATH rather than recurse
1454internally (to gain dropped root permissions or limit stack depth),
1455always call setlocale() when I18N is enabled to switch it back _off_ when
1456we run commands that expect sscanf("%n") to return bytes,
1457dirtree() had a memory leak in an error path, patch.c had some bugs in
1458error paths (didn't report problem clearly). Ashwini Sharma spotted an
1459option parsing bug where [-abc] would forget _all_ command line arguments
1460saved in the GLOBALS() block (not just the ones for options being switched
1461off), plus various minor fixes to nbd_client and cpio.
1462Lukasz Szpakowski fixed rm -f on a broken symlink (failed), and killall
1463with no arguments (segfaulted).</p>
1464
1465<p><b>Portability</b></p>
1466
1467<p>A somewhat fiddly fix to rm -rf (which needs to chmod directories to u+rwx
1468to descend into them) which hit a musl bug in faccessat() which the musl
1469maintainer refuses to fix. (He literally wants the man page changed
1470instead, despite other libcs working.) Added an #ifdef __MUSL__ section
1471to portability.h with a workaround, you may need CFLAGS=-D__MUSL__ in your
1472build if your musl build's features.h doesn't #define that. (I may do
1473a different workaround in future, but sometimes you've just got to make
1474it work so you can ship. Also, toybox grep with multiple patterns
1475requires <a href=http://landley.net/hg/aboriginal/rev/1692>a patch
1476to musl's regex engine</a>, which applies to 1.1.4 but not to the current
1477musl source control.)</p>
1478
1479<p>More portability.h fixes for uClibc too. (I don't expect that to ever have
1480another release, so locally patching around posix-2008 violations is silly).</p>
1481
1482<p><b>Change to username filtering</b></p>
1483
1484<p>Posix recommends the username creation logic filter usernames to a small
1485allowed set of characters (which even Red Hat breaks by explicitly allowing
1486"$" at the end), but this prevents UTF-8 usernames. Posix' stated logic
1487is to allow filesystems to create the user's home directory, but Linux
1488filesystems can accept any character but NUL and "/". The only characters
1489we actually _need_ to filter out are ":" (field separator in passwd),
1490newline (line separator in passwd), and "/" (directory separator in
1491filesystem).</p>
1492
1493<h3>Documentation</h3>
1494
1495<p>Web pages updated: cleanup.html documents more cleanup, code.html
1496documents more code, and about.html now capitalizes "toybox" consistently
1497(it's just a word, capitalize at start of sentence).</p>
1498
1499<p>The pending/README file now lists commands that needed review/cleanup
1500before the pending directory was added.</p>
1501
1502<h3>Test Suite</h3>
1503
1504<p>Moved out of scripts/test into top level "tests" directory, and the
1505testing.sh script is now in scripts rather than mixed into the *.test files.</p>
1506
1507<p>Johan Bergström requested VERBOSE=fail to make tests (telling it to
1508stop at the first failure), and spotted a build bug where using gnu
1509sort on the host broke in non-C locales.</p>
1510
1511<p>Divya Kothari submitted tests for chmod, link, tar, bzcat, xzcat, zcat,
1512and hostname. (And more, but that's all that's merged so far.)</p>
1513
1514<a name="07-07-2014" /><a href="#07-07-2014"><hr><h2><b>July 7, 2014</b></h2></a>
1515<blockquote><p>"This planet has - or rather had - a problem, which was this:
1516most of the people living on it were unhappy for pretty much of the time. Many
1517solutions were suggested for this problem, but most of these were largely
1518concerned with the movement of small green pieces of paper, which was odd
1519because on the whole it wasn't the small green pieces of paper that were
1520unhappy." - The Hitchhiker's Guide to the Galaxy.</p></blockquote>
1521
1522<p><a href=downloads/toybox-0.4.9.tar.bz2>Toybox 0.4.9</a> (<a href=/hg/toybox/shortlog/1385>commit 1385</a>) is out.</p>
1523
1524<p><b>New commands</b> added to pending include:
1525lsattr, chattr, inotifyd, rfkill, sulogin, strings, makedevs,
1526killall5, and tar from Ashwini Sharma, arp from Kyungwan Han,
1527sysctl by Bilal Qureshi, partprobe from Bertold Van den Bergh,
1528host from Rich felker, and I did nbd-client and the first 2/3 of mount.</p>
1529
1530<p>Finished cleanups (commands promoted out of pending):
1531sysctl, rfkill, strings, mkpasswd, makedevs, partprobe, killall5,
1532fallocate, and nbd-client.</p>
1533
1534<p>(Along the way partial cleanups got made to: last, fold, lspci, ps,
1535bootchartd, init, fsck, telnetd, telnet, vconfig, toysh, iconv, useradd,
1536login, host, openvt, deallocvt, getty, tftpd, and modprobe. But there's
1537still more to do on all of those.)</p>
1538
1539<p>This time around the <a href=bin>static binaries</a> are linked against
1540musl instead of uClibc. (That's why there's no sparc version, musl doesn't
1541support that target yet.)</p>
1542
1543<p><b>Documentation:</b></p>
1544
1545<p>The help text parser expects lower case "usage:" lines with
1546a blank line after them, so go through and regularize those. Expand the
1547"coding style" section in the docs and move it to design.html. (Not a show
1548stopper for incoming
1549contributions, just an explanation of some of the things I'll do to them
1550during cleanup.) The help text for the "toybox" command now includes
1551the shell script snippet to install symlinks to the toybox binary.</p>
1552
1553<p>The <a href=cleanup.html>cleanup page</a> now has descriptions for the
1554full ifconfig cleanup series, among others.</p>
1555
1556<p>The new toys/examples directory contains hello.c and skeleton.c. The first is
1557a simple hello world program in toybox style, the second is a much more
1558elaborate example program using showing how to use the command line option
1559parsing and how to provide multiple commands in the same C file.</p>
1560
1561<p><b>Fixes</b>:</p>
1562
1563<p>Fix od bug reported by Samuel Holland ("od -v -b" was appending the default
1564output type even though an output type was specified). Ashwini Sharma reported
1565bugs where readfile() was incorrectly freeing its buffer, and where toy_init()
1566was zeroing the wrong data because the field it was using to measure (rebound)
1567had moved (when I moved it back I added a comment why the field needs to be
1568there), fixed a segfault in the dhcp client, and made a 0 length read at
1569the start of password entry count as EOF. Make the "we are not root" test
1570in the init code show the help text. Posix implies that fflush() can return
1571success even when the stream's error bit is set, so call both fflush() and
1572ferror() from xprintf().</p>
1573
1574<p>Isaac Dunham pointed out that bloatcheck couldn't deal with diff
1575implementations that only implement "unified diff" format, and that some
1576diff implementations can't handle nonseekable input (I.E. reading from
1577a pipe). Bugfix so "help -a" works again. Option parsing on nohup now stops
1578at first nonoption argument. Fix segfault in "which" if PATH wasn't set,
1579which was actually a bug in lib function find_in_path(). Made rm -rf of
1580chmod 000 directories actually remove them.</p>
1581
1582<p>The build now passes the same $CFLAGS to the library probe as the final
1583build, because arch linux is so broken it provides different sets of
1584libraries for static and dynamic linking.</p>
1585
1586<p>It turns out sprintf("%.123s", str) is counting characters, not bytes,
1587so globally enabling locale support opens stack smashing vulnerabilities.
1588So there's a new TOYFLAGS_LOCALE you set in toyflags when you want the
1589setup code to setlocale().</p>
1590
1591<p><b>Upgrades:</b></p>
1592
1593<p>Isaac Dunham extended cpio to archive unreadable empty files, and I taught it
1594to set uid/gid and timestamp when extracting archives. Isaac also
1595added tests for cpio, link, and du, added lspci -i, made the pci database
1596parsing skip # comment lines, merged logname and whoami into id.</p>
1597
1598<p>Daniel Verkamp sped up md5sum about 30% with some loop unrolling, making
1599it actually smaller in the process. I added -b flags to md5sum and sha1sum
1600for "brief" output that's just the hash with no filename. (I'm aware other
1601implementations use that for MSDOS "binary" mode, and don't care.)</p>
1602
1603<p>When building standalone commands (scripts/singleconfig.sh commandname),
1604the build now switches on all the sub-options of the command so we get
1605a standalone version with all the bells and whistles enabled.</p>
1606
1607<p>Add -ds flags to date and document +FORMAT escapes. Add the shell NOP
1608command ":" as an alias for true (for toysh).</p>
1609
1610<p>Add uClibc probe for iconv() and fallocate. (The fact it didn't always
1611build against uClibc is why fallocate wasn't enabled in defconfig before.)</p>
1612
1613<p>The umount command now does an losetup -d on the device by default, so
1614we don't leak loopback devices. Bugfix to losetup so "losetup /dev/loop0
1615filename" actually works again.</p>
1616
1617<p>Divya Kothari sent in test suite entries for ls, ln, rm, mv, printf, dd,
1618and renice. Then a second round for lsattr/chattr, mount, chmod, pgrep/pkill,
1619groupadd, groupdel, and useradd. Several of these uncovered bugs, still
1620working to fix them.</p>
1621
1622<p>There are now free() functions for the predefined llist types and a
1623dlist_terminate() function to break doubly linked lists. The new
1624generic_signal() handler either sets "toys.signal" or writes a byte
1625to toys.signalfd with the signal number if signalfd isn't -1 (which it's
1626initialized to in toy_init).</p>
1627
1628<p>The option parsing logic can now detect when a double fits in a long and
1629use the more precise type for floating point arguments (the FLOAT macro
1630contains the type used). The human_readable() function now just outputs
1631decimal kilo/mega/gigabytes (so when du -u says 5.0G it means 5.0 billion
1632bytes). The build infrastructure now notices duplicate commands (so if you
1633cp toys/pending/command.c toys/other/command.c and forget to delete the
1634first one, the build break is now more informative).</p>
1635
1636<a name="20-04-2014" /><a href="#20-04-2014"><hr><h2><b>April 20, 2014</b></h2></a>
1637<blockquote><p>And to this end they built themselves a stupendous supercomputer
1638which was so amazingly intelligent that even before the data banks
1639had been connected up it had started from "I think therefore I am" and got as
1640far as the existence of rice pudding and income tax before anyone managed to
1641turn it off. - The Hitchhiker's Guide to the Galaxy.</p></blockquote>
1642
1643<p><a href=downloads/toybox-0.4.8.tar.bz2>Toybox 0.4.8</a> is based on
1644<a href=http://landley.net/hg/toybox/shortlog/1262>commit 1262</a>. And
1645about time too.</p>
1646
1647<p>The big news is that the build no longer needs python to generate help.h,
1648that's now done in C. The help text generation is also collating help text
1649from multiple options, merging command line option blocks and usage: lines.
1650There's even a new <a href=help.html>help web page</a>.</p>
1651
1652<p><b>New commands:</b> Ifconfig, cpio, and su were cleaned up the rest of the
1653way and promoted out of pending. That saga is mostly explained on the
1654<a href=cleanup.html>cleanup page</a>. Vivek Bhagat's freeramdisk,
1655Isaac Dunham's fsfreeze, and Felix Janda's iconv are also new.</p>
1656
1657<p><b>In pending:</b>
1658Ashwini Sharma's team submitted tcpsvd, udpsvd, telnet, telnetd, last, more,
1659groupdel/delgroup, arping, brctl, ftpget, ftpput, printf, reset, and added
1660ipv6 support to traceroute. Kyungwan Han's team submitted modprobe and getty.
1661Vivek Bhagat submitted openvt and deacllocvt. Samuel Holland submitted fold.
1662I wrote a new inflate (zip/zlib/gzip decompression) implementation in
1663compress.c, and still  need to do a corresponding deflate (compression-side)
1664and plug them into gzip and zip and so on. (Right now it does zcat.)</p>
1665
1666<p>Several commands (vmstat, login, du, vconfig, mountpoint, free, chroot,
1667cut, touch, modinfo, expand) predate the "pending" directory, and are thus
1668in other directories but still need cleanup. Of these, vmstat got some
1669work this time (which would be much easier other vmstat implementations
1670documented what their output actually meant).</p>
1671
1672<p><b>Upgrades:</b> Ifconfig grew /prefix netmask support (ala 1.2.3.4/24). Grep now has -zZ to
1673handle null terminated data, cksum grew -H for hex output. Upgraded od so the
1674fields align better when producing multiple output types. Help has -a and -h
1675options (all commands, html output).
1676Bugfix to blkid building for a 32 bit target. The date command can actually
1677set dates now. The O_NOFOLLOW compile time probe didn't work with cross
1678compiling, so it's back to an #ifdef test in portability.h. Nathan McSween
1679sent in a bugfix to od and a portability fix in the common library code.
1680Ashwini Sharma spotted a bug in pidof -o, and added verbose (-v) options
1681to mkdir and ln, and suggested killall should have an -s option and
1682allow -l to take zero arguments. Ashwini Sharma and Felix Janda upgraded
1683tftpd.  Fixed dumpleases still using toynet.h after
1684that was removed. Corrected killall return code and error reporting.
1685Isacc Dunham fixed bugs all over the tree, did cleanup on a bunch of
1686pending commands (getty, ftpget, init, openvt, modprobe...), and clarified
1687find's help text. Tom Sparrow ran three different static analyzers on
1688the code, which resulted in a few cleanups. The peek()/poke() functions
1689now use "volatile" to prevent broken compiler "optimizations" to do with
1690aliasing.</p>
1691
1692<p><b>Build stuff:</b> Each FOR_xxxx macro now has a complementary CLEANUP_xxxx macro, so you
1693can put multiple commands with different command line options in the same
1694.c file, so they can share infrastructure outside of lib. (This let the
1695bunzip logic move out of lib into bzcat.c.) See XXX for example.
1696i
1697<p>The headers #included in toys.h are now grouped by standard, and headers
1698not listed in Posix or LSB were moved to portability.h. The old xregcomp.h
1699was folded into lib.h because it's posix (and supporting oddball uClibc
1700configurations isn't as important as it once was).</p>
1701
1702<p>Regression tested against Ubuntu 8.04 to fix up bit-rot in defconfig
1703build on older systems. (We depend on Posix-2008, but not necessarily
1704the absolute latest build environment.)</p>
1705
1706<p><b>In lib</b>: lib/xwrap.c added xgetpwnam(), xchroot(), and lib/lib.c now has names_to_pid().
1707xsetuid() was replaced with xsetuser() which takes a struct passwd
1708and sets both gid and uid, mkpathat() got factored out into a library command,
1709get_int_value() became atolx_range(), and
1710xmsprintf() is now just xmprintf(). The bunzip2 logic moved from lib into
1711bzcat.c.</p>
1712
1713<p><b>Documentation</b>: new <a href=help.html>help page</a> with the
1714help text for all the defconfig commands, using the new help -ah output.
1715The <a href=code.html>source code walkthrough</a> now says more about
1716#including header files, and how the generated/* directory works. The
1717<a href=design.html>design page</a> has some new paragraphs about trading
1718of different kinds of simplicity, and why comments aren't a substitute for
1719good code. The README no longer trails off into obvious unfinished confusion
1720at the end. Each page on the website should now have its own title.</p>
1721
1722<a name="18-11-2013" /><a href="#18-11-2013"><hr><h2><b>November 18, 2013</b></h2></a>
1723<blockquote><p>"Space," it says, "is big. Really big. You just won't believe how vastly, hugely, mindbogglingly big it is. I mean, you may think it's a long way down the street to the chemist's, but that's just peanuts to space." -
1724The Hitchhiker's Guide to the Galaxy.</p></blockquote>
1725
1726<p><a href=downloads/toybox-0.4.7.tar.bz2>Toybox 0.4.7</a> is based on
1727<a href=http://landley.net/hg/toybox/shortlog/1122>commit 1122</a>.</p>
1728
1729<p>New commands: Brad Conroy submitted blkid. Elie De Brauwer submitted
1730reboot, halt, and poweroff. Strake's nl got cleaned up and promoted from
1731pending to posix. In addition, the existing chvt and vconfig got some
1732cleanup.</p>
1733
1734<p>That said, I haven't nearly kept up with the flood of new commands going
1735into pending: Ashwini Sharma's team submitted
1736dd, dumpleases, traceroute, top, useradd, groupadd, mkpasswd, tftpd, and
1737an fsck wrapper (with no filesystem drivers yet). Isaac Dunham sent in cpio.</p>
1738
1739<p>Bugfixes: Jeroen van Rijn added a user count to uptime. Elie De Brauwer
1740added -e to watch, removed a memory leak, and fixed a terminal size problem.
1741William Haddon made xargs call its command line once even with blank input
1742(the standard is vague, but builds expect it), and fixed an off by one bug
1743where grep didn't malloc enough space with -E (leading to a segfault).
1744I fixed a glitch in bunzip2 (same one as went into busybox since they're using
1745the code I wrote), in od to fix -t co, -J, and -c options. Add uname -o as a
1746synonym for -s. Build fix to never use $CC without prefixing it with
1747$CROSS_COMPILE (since $HOSTCC could be different). Anca Emanuel spotted
1748a typo in the web page.</p>
1749
1750<p>The compile-time command line option parsing got rewritten (ported from
1751bash to C), which should speed up builds a bit and allow code controlled by
1752--longopts to drop out properly when disabled in the configuration. Terminal
1753querying got refactored. Patch's -x option is now more informative (a
1754debug thing if you're trying to figure out why a patch didn't apply).
1755The "toynet.h" file got folded into toys.h since musl supports it and
1756micromanging uClibc options isn't very interesting anymore. The test suite
1757now uses scripts/single.sh when testing a single command.</p>
1758
1759<a name="17-09-2013" /><a href="#17-09-2013"><hr><h2><b>September 17, 2013</b></h2></a>
1760<blockquote><p>"Think of a number," said the computer, "any number."
1761Arthur told the computer the telephone number of King's Cross railway
1762station passenger inquiries, on the grounds that it must have some function,
1763and this might turn out to be it. - The Hitchhiker's Guide to the Galaxy</p>
1764</blockquote>
1765
1766<p><a href=downloads/toybox-0.4.6.tar.bz2>Toybox 0.4.6</a> is based on
1767<a href=http://landley.net/hg/toybox/shortlog/1068>commit 1068</a>.</p>
1768
1769<p>This release adds
1770several new commands: Felix Janda wrote paste and fallocate, Kyungwan Han
1771submitted eject, Strake contributed grep, Ashwini Sharma added pmap (and
1772a testsuite entry for grep), Lukasz Skalski sent pwdx, Isaac Dunham posted
1773acpi, and I did timeout and umount.</p>
1774
1775<p>The ls command now has a --color=auto option (suggested by Rich Felker).
1776The multiplexer now has a --help option so you can say "./toybox --help blah"
1777instead of using the built-in "help" command. (Which is a shell built-in.
1778Try it on your command line, it's like man for shell builtins. But a certain
1779other project has conditioned people to expect --help, so...) I forget who
1780heehooman at gmail is but they pointed out unshare needed PID and UID
1781namespace support.</p>
1782
1783<h3>Pending</h3>
1784
1785<p>A lot of new commands in toys/pending, to the point the next release should
1786probably just focus on cleanup and review of this backlog. We've got klogd,
1787dhcp, dhcpd, watch, route, and ps from
1788Ashwini Sharma (and an fsck wrapper but no fsck.fstype engines yet),
1789syslogd, pgrep, and pkill from Madhur Verma, netstat by Ranjan Kumar,
1790test by Felix Janda, lspci by Isaac Dunham, nl, su, and renice by strake (I.E.
1791M. Farkas-Dyck), and sysvinit by Kyungwan Han.</p>
1792
1793<p>Some cleanup work on existing pending commands that aren't
1794ready to promote yet: I did a few more rounds on ifconfig
1795and Isaac Dunham's did several cleanups to xzcat, Felix Janda cleaned up
1796logger and syslogd...</p>
1797
1798<p>Also some cleanup work on commands that predate the pending directory,
1799but weren't quite polished when they went in, most prominently du,
1800expand, and touch.</p> 
1801
1802<h3>Infrastructure</h3>
1803
1804<p>The new scripts/single.sh builds a standalone command without the
1805multiplexer, although not all commands can be built that way yet (NEWTOY yes,
1806OLDTOY no) and the space savings aren't anything to write home about. (If a
1807command needs the option parsing logic at all, it needs all of it.) If
1808you're curious, you can do:</p>
1809
1810<blockquote><pre>
1811make defconfig
1812make
1813mkdir singles
1814for i in $(./toybox)
1815do
1816  echo $i
1817  PREFIX=singles/ scripts/single.sh $i || break
1818done
1819</pre>
1820<p>(And then wait a long time and watch almost half the builds fail.)</p>
1821</blockquote>
1822
1823<p>There is now libbuf analogous to toybuf, another global 4k buffer this
1824time for use by lib/ code instead of command code.</p>
1825
1826<p>The lib directory got split up a bit, lib/pending.c contains functions
1827not yet used by anything outside of toys/pending/*, and lib/xwrap.c contains
1828functions that wrap other functions and handle failures (via error_exit).
1829This leaves lib/lib.c containing actual new functions.</p>
1830
1831<p>General improvements and bug fixes to argument parsing. The [-abc] exclude
1832logic should now clear arguments slots when disabling options. Bare --longopts
1833should work now and be able to report errors using their name, the new ;
1834option allows optional arguments to longopts only suppliable with = (I.E.
1835--color and --color=auto but not --color auto).</p>
1836
1837<p>I'm gradually weaning the code off of itoa()/utoa() because sprintf
1838does this already. In this case "simple" probably means "let libc do it
1839for us".</p>
1840
1841<p>Rewrote for_each_pid_with_name_in() and renamed it to just names_to_pid().
1842It shouldn't get confused trying to compare absolute and relative paths quite
1843so much anymore.</p>
1844
1845<p>lib/llist.c grew a new dlist_pop() function for removing a doubly
1846linked list entry while maintaining a circular list; tail and patch are
1847using it now.</p>
1848
1849<p>The musl guys suggested a new optimization flag
1850(-fno-asynchronous-unwind-tables)
1851that shaves about 10% off the binary size by removing a C++ism that crept
1852into gcc's idea of C. While I don't normally try to micromanage the compiler,
1853"-fstop-being-stupid" is a thing you have to hit gcc with from time to time.</p>
1854
1855<h3>Bugfixes</h3>
1856
1857<p>Felix Janda and I did a largeish rewrite of tail to
1858finally make it work right (we think). Still need to implement tail -f
1859someday (the tricky bit is making -f follow multiple files at once).
1860Felix also reported a bug in xpidfile.</p>
1861
1862<p>Juhani Haverinen pointed out that
1863python 3 doesn't work with config2help.py, so the detection logic looks
1864for python2 (until I get around to rewriting that in C). Elie De Brauwer
1865then fixed our first attempt at this, and also fixed uname's help string.</p>
1866
1867<p>Ashwini Sharma
1868pointed out the build was making a FLAG_ macro for " " which broke some
1869configurations. (That's a control character, not a command line option.)</p>
1870
1871<p>Jacek Bukarewicz pointed out a bug in chdir permission handling, and
1872a way to make env segfault. Both should be fixed now.</p> 
1873
1874<p>The new function xexec_optargs()
1875replaces calls to xexec(toys.optargs) to avoid freeing and reusing optargs
1876during option parsing screwing stuff up (such as netcat's exec mode).</p>
1877
1878<p>The stat command's %a output was padded with leading zeroes, which
1879didn't match anybody else's behavior and thus made the test suite hiccup
1880between TEST_HOST and testing toybox. (If you go "TEST_HOST=1 scripts/test.sh
1881command" it sanity checks the tests against the host implementation.)</p>
1882
1883<p>Last release, "mkdir sub/sub && chmod 007 sub/sub && rm -rf sub" didn't
1884delete sub and didn't exit with an error either. Neither was correct, rm
1885should now be fixed.</p>
1886
1887<p>
1888<a name="26-07-2013" /><a href="#26-07-2013"><hr><h2><b>July 26, 2013</b></h2></a>
1889<p>Georgi Chorbadzhiyski maintains a <a href=https://github.com/gfto/toybox>git
1890mirror</a> of the repository on github, automatically updated from the
1891mercurial every 6 hours. The mirror is read only, but you can generate patches
1892against it and post them to the list.</p>
1893
1894<a name="02-07-2013" /><a href="#02-07-2013"><hr><h2><b>July 2, 2013</b></h2></a>
1895<blockquote><p>"Time is an illusion. Lunchtime doubly so." "Very deep. You
1896should send that in to the Reader's Digest. They've got a page for people
1897like you." -
1898The Hitchhiker's Guide to the Galaxy.</p></blockquote>
1899
1900<p><a href=downloads/toybox-0.4.5.tar.bz2>Toybox 0.4.5</a> is based on
1901<a href=http://landley.net/hg/toybox/shortlog/941>commit 941</a>. It adds
1902uuencode and uudecode from Erich Plondke, and enables Luis Morales' "who" by
1903default. Felix Janda and I cleaned up last year's "stat" submission and
1904enabled it. Ivo van Poorten added "groups".
1905Andre Renaud added "lsusb". I implemented "split", "pivot_root", and "mv".
1906</p>
1907
1908<p>The "help" command is implemented differently now (lib/help.c) and
1909each command can now understand --help (including both "toybox --help"
1910and "toybox --help command" in the multiplexer).</p>
1911
1912<p>The "pending" directory has several commands (find, xzcat, nbd-client,
1913logger, expr) which work but are not enabled by default pending further cleanup.
1914Ifconfig is enabled, but still in pending because it's only 2/3 cleaned up.
1915(It's an awkward halfway state but I'm not holding up the release for it.)</p>
1916
1917<p>I'm <a href=cleanup.html>documenting the cleanups</a> to teach
1918more people to do it, but the writeups aren't caught up yet. The
1919<a href=roadmap.html>roadmap</a> also got updated a bit with further analysis
1920of other projects, and the README and about pages got updated.</p>
1921
1922<p>Fixed _another_ "ls -C" segfault when terminal size can't be detected,
1923condensed the ls help text to fit on one page, implented --color, and taught
1924-l to print the major, minor numbers when showing block/char devices.
1925Argument parsing now handles "--" properly (to end option checking),
1926and the infrastructure can now handle bare --longopts that have no
1927corresponding short option (both were implemented before but didn't work).
1928Fixed an old bug in "patch", chmod grew -f, who grew -a. Isaac Dunham
1929fixed "-" vs "_" handling in modinfo, added a "firmware" output
1930field, added -b and -k support, and taught it that the ".ko" extension means
1931to look for the file at the specified path instead of under /lib. Felix Janda
1932moved file permission display code to lib so ls and
1933stat could share it. Ashwini Sharma spotted a bug in xabspath when the
1934last path component exists but we haven't got permissions to open it
1935(ala readlink -f /dev/sda as a normal user).
1936</p>
1937
1938<p>In the build infrastructure, scripts/findglobals.sh finds leaked global
1939variables. (Leaked means they aren't part of the global union: Other than glibc
1940debris, toybox should define "this", "toy_list", "toybuf", and "toys", and
1941that's it; the rest add memory footprint to every command for the benefit of
1942just one command; use GLOBALS() to stick 'em in the union.) Static linking
1943against libraries other than the host's libc now applies to feature probes
1944for unshare and such. Neuter stupid internationalization support that makes
1945various host "sort" commands put things in an order other than alphabetical
1946(breaking the multiplexer's binary search on command names).
1947
1948<p>You should now be able to build from a source control snapshot on a build
1949system that hasn't got python: if you disable CONFIG_TOYBOX_HELP. (The
1950release tarballs ship generated/help.h, but it's not in source control.
1951Eventually I should rewrite that python script in C.)</p>
1952</p>
1953
1954<p><b>LICENSE TWEAK</b>: After <a href=http://lists.landley.net/pipermail/toybox-landley.net/2013-March/000794.html>discussion</a> on the mailing list the "2 clause
1955BSD" <a href=license.html>license</a> got slightly simplified so the first
1956paragraph now says:</p>
1957
1958<blockquote><p>Permission to use, copy, modify, and/or distribute this
1959software for any purpose with or without fee is hereby granted.</p></blockquote>
1960
1961<p>It used to continue "provided that the above copyright notice and this
1962permission notice appear in all copies", but A) what's the point? B) does "all
1963copies" mean binaries, or just source code, or what? C) lots of projects
1964that consider BSD and GPL compatible have <a href=https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/crypto/aes_generic.c>files with
1965both license notices</a> on them (sometimes at <a href=http://git.busybox.net/busybox/tree/shell/ash.c>opposite ends of the file</a> to make the conflict
1966less obvious) because "all copies must include this function" would violate
1967the GPL but "all copies must include this magic text blob" somehow don't?</p>
1968
1969<p>I don't want to have to care about this anymore. The tweaked version is more
1970or less public domain with a liability disclaimer, but we're still calling it
1971BSD (sometimes "0 clause BSD") to avoid explaining.</p>
1972
1973<a name="21-03-2013" /><a href="#21-03-2013"><hr><h2><b>March 21, 2013</b></h2></a>
1974<p>Video of my ELC talk
1975"<a href=http://youtu.be/SGmtP5Lg_t0>Why is Toybox?</a>"
1976is up on youtube. Related materials include the
1977<a href=http://landley.net/talks/celf-2013.txt>talk outline</a> and an
1978<a href=/aboriginal/about.html#selfhost>android self-hosting writeup</a>.</p>
1979
1980<p>[Updated June 4] The following links jump to specific topics in the video. (Sorry about
1981the ads, it's The Linux Foundation.)</p>
1982
1983<ul>
1984<li>0m29s <a href=http://www.youtube.com/watch?v=SGmtP5Lg_t0#t=0m29s>The smartphone is replacing the PC</a></li>
1985  <ul>
1986  <li>4m22s <a href=http://www.youtube.com/watch?v=SGmtP5Lg_t0#t=4m22s>Software needed to become self-hosting</a></li>
1987  <li>6m20s <a href=http://www.youtube.com/watch?v=SGmtP5Lg_t0#t=6m20s>Do we care if android or iphone wins?</a></li>
1988  </ul>
1989<li>9m45s <a href=http://www.youtube.com/watch?v=SGmtP5Lg_t0#t=9m45s>Android not vanilla: oppose or accept?</a></li>
1990  <ul>
1991  <li>11m30s <a href=http://www.youtube.com/watch?v=SGmtP5Lg_t0#t=11m30s>Open source can't do User Interfaces</a></li>
1992  </ul>
1993<li>15m09s <a href=http://www.youtube.com/watch?v=SGmtP5Lg_t0#t=15m09s>Android is not copyleft: oppose or accept?</a></li>
1994<li>18m23s <a href=http://www.youtube.com/watch?v=SGmtP5Lg_t0#t=18m23s>Security issues</a></li>
1995<li>21m15s <a href=http://www.youtube.com/watch?v=SGmtP5Lg_t0#t=21m15s>Solutions to the software problems</a></li>
1996  <ul>
1997  <li>22m55s <a href=http://www.youtube.com/watch?v=SGmtP5Lg_t0#t=22m55s>What toybox needs to be/do</a></li>
1998  <li>28m17s <a href=http://www.youtube.com/watch?v=SGmtP5Lg_t0#t=28m17s>What is toybox?</a></li>
1999    <ul>
2000    <li>28m58s <a href=http://www.youtube.com/watch?v=SGmtP5Lg_t0#t=28m58s>Why toybox started...</a></li>
2001    <li>37m50s <a href=http://www.youtube.com/watch?v=SGmtP5Lg_t0#t=37m50s>What does toybox actually implement?</a></li>
2002    </ul>
2003  </ul>
2004</ul>
2005</span>
2006
2007
2008<a name="14-03-2013" /><a href="#14-03-2013"><hr><h2><b>March 14, 2013</b></h2></a>
2009<blockquote><p>"Ford, you're turning into a penguin. Stop it." -
2010The Hitchhiker's Guide to the Galaxy.</p></blockquote>
2011
2012<p><a href=downloads/toybox-0.4.4.tar.bz2>Toybox 0.4.4</a> is based on
2013<a href=http://landley.net/hg/toybox/shortlog/813>commit 813</a>, adding
2014the "time" and "readahead" commands, plus some bugfixes.</p>
2015
2016<p>The "cp" command now implements the -s symlink option, plus bugfixes
2017getting various corner cases right as used in actual package builds.
2018"id -Gn root" should now print root's groups
2019instead of the current user's. Several build fixes so toybox builds under
2020Ubuntu 8.04 again (which is about as old a build environment as you
2021can expect to find posix-2008 features in).</p>
2022
2023<p>Unfinished commands have generally been moved to "toys/pending".
2024Everything else should "default y" to participate in make defconfig.
2025Several of those pending commands got some basic cleanup so allyesconfig
2026should at least compile (although defconfig is still what's useful).</p>
2027
2028<p>Significant roadmap updates, checking several other multicall binaries
2029(klibc, sash, sbase, s6...) to see what commands they include.</p>
2030
2031<a name="18-01-2013" /><a href="#18-01-2013"><hr><h2><b>January 18, 2013</b></h2></a>
2032<blockquote><p>This must be Thursday. I never could get the hang of Thursdays. - The Hitchhiker's Guide to the Galaxy.</p></blockquote>
2033
2034<p><a href=downloads/toybox-0.4.3.tar.bz2>Toybox 0.4.3</a> is based on
2035<a href=http://landley.net/hg/toybox/shortlog/793>commit 793</a>. There
2036are now exactly 100 commands in defconfig (of a little over 220 on the
2037<a href=roadmap.html>todo list</a>).</p>
2038
2039<p>Elie De Brauwer added the rev command, cleaned up tac, implemented the -s
2040and -f flags for seq, added -v and -i to killall (and fixed killall not to
2041kill itself before finishing its pid list), and added to the test suite.
2042Felix Janda added -m to mkdir, pwd -L and -P, and more test suite entries.</p>
2043
2044<p>Rob Landley added the losetup command, and fixed the existing ls, cp, and
2045readlink commands. The segfault in ls
2046happened when it couldn't determine the screen size (last release changed the
2047default to -C and a screen size of 0 made column view unhappy), and cp got an
2048extensive rewrite bringing it up to date with the dirtree changes and fixing
2049a number of things it never did right in the first place. The xabspath()
2050code in the library now handles a symlink after ".." properly (and the test
2051suite checks for it).</p>
2052
2053<p>Infrastructure-wise the code is better about automatically setting the
2054error return code properly. Now error_msg() sets the exit code to 1 if it's
2055still defaulting to 0, and the global exit path does a fflush(NULL) with error
2056bit check rather than trying to be quite so granular about flushing. (That
2057means if we use printf() instead of xprintf() it still exits with the right
2058error code, it just doesn't end the program early on an output error.)
2059Minor bugfix so TOYBOX_DEBUG
2060doesn't always warn about the lack of suid bit when toybox is built with
2061at least one STAYROOT command. Bugfix for the option [grouping] logic
2062(and then further fixes to the error reporting pointed out by Ashwini Sharma).
2063dirtree_handle_callback() now has a prefix like the rest of the dirtree
2064functions. A lot of stuff doing manual path handling was switched to using
2065libc basename() (including, embarassingly, the basename command), which means
2066it now correctly detects "/trailing/slash/" which the previous code didn't.</p>
2067
2068<p>Also, last release included some accidentally checked in debug code that
2069disabled compiler optimization, so the binary size bloated a bit. It's back
2070to -Os by default now.</p>
2071
2072<a name="15-12-2012" /><a href="#15-12-2012"><hr><h2><b>December 15, 2012</b></h2></a>
2073<blockquote><p>"The major difference between a thing that might go wrong and a
2074thing that cannot possibly go wrong is that when a thing that cannot possibly
2075go wrong goes wrong it usually turns out to be impossible to get at or repair."
2076</p><p>- The Hitchhiker's Guide to the Galaxy.</p></blockquote>
2077
2078<p><a href=downloads/toybox-0.4.2.tar.bz2>Toybox 0.4.2</a> is based on
2079<a href=http://landley.net/hg/toybox/shortlog/749>commit 749</a> and is
2080just a resync. Linux 3.7 came out, meaning it's time to do an Aboriginal
2081Linux release, and that should use a stable version of toybox. So here's
2082a new stable version.</p>
2083
2084<p>The new commands are cut (from Jason Kyungwan Han), touch
2085(from Choubey Ji), expand (from Jonathan Clairembault, and he fixed a
2086bug in login), and rm (from Rob Landley). Felix Janda added UTF-8
2087support infrastructure (for non-ascii character sets) with a config option.
2088Elie De Brauwer added tests for cat and sha1sum, and -so options to pidof.
2089The "ls" command defaults to -C (column view) now, and "readlink" now supports
2090-fenq.</p>
2091
2092<p>Portability work: toybox should now build against the musl C library,
2093and against older glibc versions (circa 2008, much before that and kernel
2094features we depend on start to drop out).</p>
2095
2096<p>The whole codebase got reindented from "one tab" to "two spaces" per
2097level. The option parsing logic now understands [groups] of commands (when more
2098than one in a group is selected it can switch the others off, or error out,
2099or other things). The error_exit() infrastructure can now longjmp back to an
2100earlier point instead of exiting. Each toys/* directory now has a README,
2101the first line of which is the fancy name menuconfig uses for the directory
2102(so no more hardwired directory list in scripts/genconfig.sh).</p>
2103
2104<p>Fixed a filehandle leak in getmountlist().
2105Pass parent pointer to dirtree_add_node() so it can give error messages with
2106full path. The yesno() function now always reads from stdin and writes to
2107stderr (we can retry tty checking complexity once we've got commands needing
2108it).</p>
2109
2110<p>The open group broke their website so the
2111<a href=http://opengroup.org/onlinepubs/9699919799>old links</a> to POSIX 2008
2112now <a href=http://pubs.opengroup.org/onlinepubs/9699919799>need to start with
2113pubs</a>. Some of the links in the tree have been updated, others haven't while
2114I wait to see if their webmaster notices and fixes it.</p>
2115
2116<p>(I note that the current rm implementation is not technically posix compliant
2117because the standard requires infinite recursion depth and the current
2118implementation uses one filehandle per level. I can add a config option
2119to do it Posix's way, which is more brittle and needs extra security checks,
2120but am waiting for somebody to complain first. The default "ulimit -n" is 1024
2121filehandles, so drilling down over 1000 nested subdirectories).</p>
2122
2123<a name="13-11-2012" /><a href="#13-11-2012"><hr><h2><b>November 13, 2012</b></h2></a>
2124<blockquote><p>"Rule Six: The winning team shall be the first team that wins."
2125- The Hitchhiker's Guide to the Galaxy.</p></blockquote>
2126
2127<p><a href=downloads/toybox-0.4.1.tar.bz2>Toybox 0.4.1</a> is based on
2128<a href=http://landley.net/hg/toybox/shortlog/691>commit 691</a>.</p>
2129
2130<p>Elie De Brauwer contributed usleep, Ashwini Kumar contributed du, and
2131Kyungwan Han contributed vconfig. Other new commands include switch_root and
2132md5sum, and the remaining shell wrappers are now proper commands (dos2unix,
2133unix2dos).</p>
2134
2135<p>The patch command now supports -l, and gethostname is now enabled by
2136default. The df command follows symlinks to get the actual device name.
2137Felix Janda added -m support to wc (for utf8).</p>
2138
2139<p>On the infrastructure side, the commands have now been grouped into
2140"posix", "lsb", and "other" subdirectories (for things required by Posix-2008,
2141the Linux Standard Base 4.1, and commands in neither). This affects menuconfig
2142and the actual source layout (toys/cp.c is now toys/posix/cp.c, and so on).
2143An android directory is planned (see the updated
2144<a href=roadmap.html#android>android roadmap analysis</a>).</p>
2145
2146<p>The FLAG_ macros for command option parsing and TT alias for the command's
2147global block are now automatically generated, commands should
2148#define FOR_commandname before #including <toys.h> to get the macros for that
2149command.</p>
2150
2151<p>An upgrade to the build infrastructure now allows commands with _ and -
2152in them, such as switch_root.</p>
2153
2154<p>Bugfixes: Avery Pennarun spotted a case where ls showed uid twice instead of
2155uid and gid, and that nice was using the wrong range of numbers.
2156The ls command also recursed inappropriately last time (not quite
2157properly converted for the dirtree changes last release), and now it's
2158fixed. Roy Tam pointed out a glitch in sh, and fixed df's percentage
2159calculation to match the POSIX spec. The kernel build didn't like our mktemp
2160and it does now. The wc command wasn't quite posix compliant (trailing spaces
2161break stuff). The ls command recursed inappropriately last time (not quite
2162properly converted for the dirtree changes last release), and now it's
2163fixed. The catv command wasn't displaying byte 255 correctly. Some lib
2164fixes (thinko in xpidfile). Fixed uname -m when running a 32 bit x86 binary
2165on an x86-64 host (it lies and says the system is i686, i586, or i486 depending
2166on what the toolchain that built the binary supported. This makes builds in
2167a 32 bit chroot on a 64 bit kernel break less.) The df command was checking
2168partitions in the wrong order (displaying undermounts instead of overmounts:
2169this used to work but some library code changed out from under it and it
2170wasn't updated to match until now). Felix Janda filled out the test suite
2171some more. The patch file creation logic got tweaked several times to
2172successfully apply more patches. Support for older (pre 2.10) glibc
2173versions was added to portability.h.</p>
2174
2175<p>Miscelaneous cleanups all around (mknod, sha1sum, logname), including a
2176rewrite of taskset to be less dependent on libc getting the headers right. All
2177the command headers should now point to the current relevant standards
2178document, where applicable.</p>
2179
2180<p>This news page had old news entries from before the relaunch moved into
2181a separate <a href=oldnews.html>oldnews</a> page.</p>
2182
2183<p>I forgot to create <a href=bin>static binaries</a> last time, but they're
2184back now.</p>
2185</span>
2186
2187<a name="23-07-2012" /><a href="#23-07-2012"><hr><h2><b>July 23, 2012</b></h2></a>
2188<blockquote><p>"Ford", Arthur said. "There's an infinite number of monkeys
2189out here who want to talk to us about this script for Hamlet they've worked
2190out." - The Hitchhiker's Guide to the Galaxy.</p></blockquote>
2191
2192<p><a href=downloads/toybox-0.4.0.tar.bz2>Toybox 0.4.0</a> is based on
2193<a href=http://landley.net/hg/toybox/shortlog/640>commit 640</a>.</p>
2194
2195<p>The new <a href=status.html>status page</a> is calculated from
2196the roadmap info, and should be easier to keep up to date in future.</p>
2197
2198<p>Andre Renaud contributed od and modinfo. Elie De Brauwer contributed
2199taskset, bugfixes to cmp and tail, and tests for sort and tail. Kyungwan Han
2200contributed passwd. Gaurang Shastri contributed w. Ashwini Sharma spotted a
2201case where dirtree was adding extra slashes to a path.</p>
2202
2203<p>I rewrote od, cleaned up comm, documented the
2204<a href=code.html#lib_llist>llist</a> and
2205<a href=code.html#lib_dirtree>dirtree</a> infrastructure, added an -r option
2206to date (and fixed a bug where -u wouldn't override /etc/localtime),
2207fixed bugs in chmod +stw, fixed ls to show suid bits properly when the
2208corresponding executable bit wasn't set, and worked around a longstanding
2209glibc bug where static linking prevents stdout from automatically flushing
2210pending output on exit.</p>
2211
2212<a name="25-06-2012" /><a href="#25-06-2012"><hr><h2><b>June 25, 2012</b></h2></a>
2213<blockquote><p>"For a moment, nothing happened. Then, after a second or so, nothing continued to happen." - The Hitchhiker's Guide to the Galaxy.</p></blockquote>
2214
2215<p><a href=downloads/toybox-0.3.1.tar.bz2>Toybox 0.3.1</a> is based on commit
2216<a href=http://landley.net/hg/toybox/shortlog/607>commit 607</a>. It's
2217mostly a bugfix release for ls -l (which was unhappy on targets other than
2218x86-64), plus a new "date" from Andre Renaud and rewritten chgrp/chown which
2219now support the full set of posix flags, plus a little work on the test
2220suite and some more header tweaks towards eventual compatability with the
2221musl libc.</p>
2222
2223<p>The todo list runneth over, but "release early, release often", so here
2224it is. The roadmap and documentation are a bit behind, and I've got ~40
2225pending submissions to review. I need to catch up...</p>
2226</span>
2227
2228<a name="12-06-2012" /><a href="#12-06-2012"><hr><h2><b>June 12, 2012</b></h2></a>
2229<blockquote><p>"For instance, on the planet Earth, man had always assumed that
2230he was more intelligent than dolphins because he had achieved so much - the
2231wheel, New York, wars and so on - whilst all the dolphins had ever done was
2232muck about in the water having a good time. But conversely, the dolphins had
2233always believed that they were far more intelligent than man - for precisely
2234the same reasons." - The Hitchhiker's Guide to the Galaxy.</p></blockquote>
2235
2236<p>It's well past time for <a href=downloads/toybox-0.3.0.tar.bz2>toybox 0.3.0</a>,
2237so here it is, based
2238on <a href=http://landley.net/hg/toybox/shortlog/595>commit 595</a>, and the
2239statically linked <a href=downloads/binaries>prebuilt binaries</a> should
2240actually be statically linked this time (thanks Ashwini Sharma for spotting
2241that).</p>
2242
2243<p>It's hard to figure out where to cut a release, because development
2244doesn't stop. "Long before now" is the obviuos answer, of course.
2245The project's maintainer also moved house during this development cycle, which
2246threw things off for a bit (so many boxes). Releases should hopefully be a bit
2247more frequent from here on.</p>
2248
2249<p>The big things Rob worked on this time were the new dirtree (directory
2250tree traversal) infrastructure, and a complete rewrite of ls using that
2251which should now implement all 26 posix options.</p>
2252
2253<p>Georgi Chorbadzhiyski added printenv, whoami, mkdir, mkfifo, chmod, chown,
2254chgrp, and uniq. He also added fraction and extension support to sleep (so if
2255you need a quarter-second sleep, it can do that now), and fixed a build bug
2256on slackware.</p>
2257
2258<p>Daniel Walter contributed a string to mode_t parser (in use by chmod and
2259mkdir -m).  Ilya Kuzmich contributed comm. Elie De Brauwer added mountpoint,
2260vmstat, logname, login, and mktemp. Kevin Chase did some portability cleanups.
2261Pere Orga fixed some documentation.</p>
2262
2263<p>The "tac" and "clear" commands are now normal commands instead of shell
2264wrappers, and the header #includes have been cleaned up a bit to remove
2265deprecated functions and attempt to increase compatability with the bionic and
2266musl C libraries, "tail" should now use lseek() for large files, and "id" got
2267some cleanups and bugfixes.</p>
2268
2269<p>The new TOYBOX_FLOAT configuration option selects whether or not
2270to include floating point support (for embedded targets where that's
2271problematic).</p>
2272
2273<p>Several random bugfixes: unshare() might actually build portably now,
2274yes 'n' | cp -i should no longer bypass stdin and prompt via the tty, the
2275SUID support no longer drops permissions going through the toybox
2276multiplexer command, and a bugfix to xargs -0 means it should no longer
2277segfault. (I have a pending bug report about xargs not doing the full
2278posix whitespace handling that -0 obsoleted, but I'll deal with that next
2279release.)</p>
2280
2281<p>The build infrastructure is now automatically generating FLAG_ macros
2282for the options, but currently with the wrong names. Some more macro glue
2283is necessary, which I haven't quite figured out how to do yet.</p>
2284
2285<p>A defconfig toybox at the start of the $PATH has successfully built
2286Linux From Scratch (in my Aboriginal Linux project). The commands that
2287'default n' in the config are often still broken, cleanup is ongoing.
2288(The new dirtree stuff broke several of them that haven't been converted
2289yet, but if I wait until everything works we won't have a release before
22901.0, so here's a checkpoint.)</p>
2291
2292
2293<a name="03-03-2012" /><a href="#03-03-2012"><hr><h2><b>March 3, 2012</b></h2></a>
2294
2295<blockquote><p>"They went unnoticed at Goonhilly, passed over Cape Canaveral
2296without a blip, and Woomera and Jodrell Bank looked straight through them.
2297Which was a pity, because it was exactly the sort of thing they'd been looking
2298for all these years."</p></p>- The Hitchhiker's Guide to the Galaxy.</p>
2299</p></blockquote>
2300
2301<p>Here's <a href=downloads/toybox-0.2.1.tar.bz2>toybox 0.2.1</a> based
2302on <a href=http://landley.net/hg/toybox/shortlog/512>commit 512</a>.  This
2303time around, there are statically linked <a href=downloads/binaries>prebuilt
2304binaries</a> for various embedded targets.</p>
2305
2306<p>It's been a busy few weeks, almost entirely due to new contributors. (I
2307have not quite been keeping up.)</p>
2308
2309<p>Elie De Brauwer contributed free, uptime, swapon, swapoff, lsmod, mknod,
2310insmod, rmmod, and fixed a bug in basename.  Andre Renaud contributed ls, ln,
2311realpath, and hostname. Andres Heck contributed pidof and killall.  Daniel
2312Walter wrote kill and extended id. Timothy Elliott contributed tail and tests
2313for cmp. Frank Bergmann sent a warning fix. Bryce Fricke added -i to cp.
2314Nathan McSween pointed out an optimization. Georgi Chorbadzhiyski fixed
2315cross compiling to work more reliably.</p>
2316
2317<p>(My own contribution this time around was just tightening up other people's
2318code, a build fix to unshare, some random bugfixes, and so on. My only new
2319code this time around was writing a bash replacement for the existing python
2320bloat-o-meter.)</p>
2321
2322<p>Last time (the 0.2.0 release) included the first pass at an id command from
2323Tim Bird, env and basename from Tryn Mirell, cmp and head from Timothy Elliott,
2324more bugfixes from Nathan McSween and Elie De Brauwer, and Luis Felipe Strano
2325Moraes did a first pass at the who command plus other bugfixes and
2326optimizations.</p>
2327
2328<p>(For that release I did xargs, cal, truncate, unlink, nohup, tty, wc, link,
2329dirname, unshare, and various infrastructure tweaks, but it took me 3 months
2330and those guys did their stuff in a week or so.)</p>
2331
2332
2333<a name="12-02-2012" /><a href="#12-02-2012"><hr><h2><b>February 12, 2012</b></h2></a>
2334<blockquote><p>
2335"for though it has many omissions and contains much that is apocryphal, or at
2336least wildly inaccurate, it scores over the older, more pedestrian work in two
2337important respects..."</p>
2338<p> - The Hitchhiker's Guide to the Galaxy</p></blockquote>
2339
2340<p>Here's the first BSD licensed release,
2341<a href=downloads/toybox-0.2.0.tar.bz2>toybox-0.2.0</a>, more a synchronization
2342point than anything particularly useful.  47 commands in a reasonably
2343ready-to-use state (what "make defconfig" builds), another ten or so partially
2344finished stubs ("make allyesconfig"), and several
2345patches pending on the mailing list I need to review and merge.</p>
2346
2347<p>More to come...</p>
2348
2349<hr>
2350<a name="15-11-2011" /><a href="#15-11-2011"><hr><h2><b>November 15, 2011</b></h2></a>
2351- Back from the dead, Toybox is now under a 2
2352clause BSD license, and aiming to become the default command line
2353implementation of Android systems everywhere.</p>
2354
2355<p>More to come...</p>
2356
2357<hr>
2358
2359<p><a href=oldnews.html>Old news</a> from before the relaunch.</p>
2360
2361<!--#include file="footer.html" -->
2362