toybox/www/code.html
<<
>>
Prefs
   1<html><head><title>toybox source code walkthrough</title></head>
   2<!--#include file="header.html" -->
   3
   4<p><h1><a name="style" /><a href="#style">Code style</a></h1></p>
   5
   6<p>The primary goal of toybox is _simple_ code. Keeping the code small is
   7second, with speed and lots of features coming in somewhere after that.
   8(For more on that, see the <a href=design.html>design</a> page.)</p>
   9
  10<p>A simple implementation usually takes up fewer lines of source code,
  11meaning more code can fit on the screen at once, meaning the programmer can
  12see more of it on the screen and thus keep more if in their head at once.
  13This helps code auditing and thus reduces bugs. That said, sometimes being
  14more explicit is preferable to being clever enough to outsmart yourself:
  15don't be so terse your code is unreadable.</p>
  16
  17<p>Toybox has an actual coding style guide over on
  18<a href=design.html#codestyle>the design page</a>, but in general we just
  19want the code to be consistent.</p>
  20
  21<p><h1><a name="building" /><a href="#building">Building Toybox</a></h1></p>
  22
  23<p>Toybox is configured using the Kconfig language pioneered by the Linux
  24kernel, and adopted by many other projects (uClibc, OpenEmbedded, etc).
  25This generates a ".config" file containing the selected options, which
  26controls which features are included when compiling toybox.</p>
  27
  28<p>Each configuration option has a default value. The defaults indicate the
  29"maximum sane configuration", I.E. if the feature defaults to "n" then it
  30either isn't complete or is a special-purpose option (such as debugging
  31code) that isn't intended for general purpose use.</p>
  32
  33<p>For a more compact human-editable version .config files, you can use the
  34<a href=http://landley.net/aboriginal/FAQ.html#dev_miniconfig>miniconfig</a>
  35format.</p>
  36
  37<p>The standard build invocation is:</p>
  38
  39<ul>
  40<li>make defconfig #(or menuconfig)</li>
  41<li>make</li>
  42<li>make install</li>
  43</ul>
  44
  45<p>Type "make help" to see all available build options.</p>
  46
  47<p>The file "configure" contains a number of environment variable definitions
  48which influence the build, such as specifying which compiler to use or where
  49to install the resulting binaries. This file is included by the build, but
  50accepts existing definitions of the environment variables, so it may be sourced
  51or modified by the developer before building and the definitions exported
  52to the environment will take precedence.</p>
  53
  54<p>(To clarify: ".config" lists the features selected by defconfig/menuconfig,
  55I.E. "what to build", and "configure" describes the build and installation
  56environment, I.E. "how to build it".)</p>
  57
  58<p>By default "make install" puts files in /usr/toybox. Adding this to the
  59$PATH is up to you. The environment variable $PREFIX can change the
  60install location, ala "PREFIX=/usr/local/bin make install".</p>
  61
  62<p>If you need an unstripped (debug) version of any of these binaries,
  63look in generated/unstripped.</p>
  64
  65<p><h1><a name="running"><a href="#running">Running a command</a></h1></p>
  66
  67<h2>main</h2>
  68
  69<p>The toybox main() function is at the end of main.c at the top level. It has
  70two possible codepaths, only one of which is configured into any given build
  71of toybox.</p>
  72
  73<p>If CONFIG_SINGLE is selected, toybox is configured to contain only a single
  74command, so most of the normal setup can be skipped. In this case the
  75multiplexer isn't used, instead main() calls toy_singleinit() (also in main.c)
  76to set up global state and parse command line arguments, calls the command's
  77main function out of toy_list (in the CONFIG_SINGLE case the array has a single entry, no need to search), and if the function returns instead of exiting
  78it flushes stdout (detecting error) and returns toys.exitval.</p>
  79
  80<p>When CONFIG_SINGLE is not selected, main() uses basename() to find the
  81name it was run as, shifts its argument list one to the right so it lines up
  82with where the multiplexer function expects it, and calls toybox_main(). This
  83leverages the multiplexer command's infrastructure to find and run the
  84appropriate command. (A command name starting with "toybox" will
  85recursively call toybox_main(); you can go "./toybox toybox toybox toybox ls"
  86if you want to...)</p>
  87
  88<h2>toybox_main</h2>
  89
  90<p>The toybox_main() function is also in main,c. It handles a possible
  91--help option ("toybox --help ls"), prints the list of available commands if no
  92arguments were provided to the multiplexer (or with full path names if any
  93other option is provided before a command name, ala "toybox --list").
  94Otherwise it calls toy_exec() on its argument list.</p>
  95
  96<p>Note that the multiplexer is the first entry in toy_list (the rest of the
  97list is sorted alphabetically to allow binary search), so toybox_main can
  98cheat and just grab the first entry to quickly set up its context without
  99searching. Since all command names go through the multiplexer at least once
 100in the non-TOYBOX_SINGLE case, this avoids a redundant search of
 101the list.</p>
 102
 103<p>The toy_exec() function is also in main.c. It performs toy_find() to
 104perform a binary search on the toy_list array to look up the command's
 105entry by name and saves it in the global variable which, calls toy_init()
 106to parse command line arguments and set up global state (using which->options),
 107and calls the appropriate command's main() function (which->toy_main). On
 108return it flushes all pending ansi FILE * I/O, detects if stdout had an
 109error, and then calls xexit() (which uses toys.exitval).</p>
 110
 111<p><h1><a name="infrastructure" /><a href="#infrastructure">Infrastructure</a></h1></p>
 112
 113<p>The toybox source code is in following directories:</p>
 114<ul>
 115<li>The <a href="#top">top level directory</a> contains the file main.c (were
 116execution starts), the header file toys.h (included by every command), and
 117other global infrastructure.</li>
 118<li>The <a href="#lib">lib directory</a> contains common functions shared by
 119multiple commands:</li>
 120<ul>
 121<li><a href="#lib_lib">lib/lib.c</a></li>
 122<li><a href="#lib_xwrap">lib/xwrap.c</a></li>
 123<li><a href="#lib_llist">lib/llist.c</a></li>
 124<li><a href="#lib_args">lib/args.c</a></li>
 125<li><a href="#lib_dirtree">lib/dirtree.c</a></li>
 126</ul>
 127<li>The <a href="#toys">toys directory</a> contains the C files implementating
 128each command. Currently it contains five subdirectories categorizing the
 129commands: posix, lsb, other, example, and pending.</li>
 130<li>The <a href="#scripts">scripts directory</a> contains the build and
 131test infrastructure.</li>
 132<li>The <a href="#kconfig">kconfig directory</a> contains the configuration
 133infrastructure implementing menuconfig (copied from the Linux kernel).</li>
 134<li>The <a href="#generated">generated directory</a> contains intermediate
 135files generated from other parts of the source code.</li>
 136<li>The <a href="#tests">tests directory</a> contains the test suite.
 137NOSPACE=1 to allow tests to pass with diff -b</li>
 138</ul>
 139
 140<a name="adding" />
 141<p><h1><a href="#adding">Adding a new command</a></h1></p>
 142<p>To add a new command to toybox, add a C file implementing that command to
 143one of the subdirectories under the toys directory.  No other files need to
 144be modified; the build extracts all the information it needs (such as command
 145line arguments) from specially formatted comments and macros in the C file.
 146(See the description of the <a href="#generated">"generated" directory</a>
 147for details.)</p>
 148
 149<p>Currently there are five subdirectories under "toys", one for commands
 150defined by the POSIX standard, one for commands defined by the Linux Standard
 151Base, an "other" directory for commands not covered by an obvious standard,
 152a directory of example commands (templates to use when starting new commands),
 153and a "pending" directory of commands that need further review/cleanup
 154before moving to one of the other directories (run these at your own risk,
 155cleanup patches welcome).
 156These directories are just for developer convenience sorting the commands,
 157the directories are otherwise functionally identical. To add a new category,
 158create the appropriate directory with a README file in it whose first line
 159is the description menuconfig should use for the directory.)</p>
 160
 161<p>An easy way to start a new command is copy the file "toys/example/hello.c"
 162to the name of the new command, and modify this copy to implement the new
 163command (more or less by turning every instance of "hello" into the
 164name of your command, updating the command line arguments, globals, and
 165help data, and then filling out its "main" function with code that does
 166something interesting).</p> 
 167
 168<p>You could also start with "toys/example/skeleton.c", which provides a lot
 169more example code (showing several variants of command line option
 170parsing, how to implement multiple commands in the same file, and so on).
 171But usually it's just more stuff to delete.</p>
 172
 173<p>Here's a checklist of steps to turn hello.c into another command:</p>
 174
 175<ul>
 176<li><p>First "cp toys/example/hello.c toys/other/yourcommand.c" and open
 177the new file in your preferred text editor.</p>
 178<ul><li><p>Note that the
 179name of the new file is significant: it's the name of the new command you're
 180adding to toybox. The build includes all *.c files under toys/*/ whose
 181names are a case insensitive match for an enabled config symbol. So
 182toys/posix/cat.c only gets included if you have "CAT=y" in ".config".</p></li>
 183</ul></p></li>
 184
 185<li><p>Change the one line comment at the top of the file (currently
 186"hello.c - A hello world program") to describe your new file.</p></li>
 187
 188<li><p>Change the copyright notice to your name, email, and the current
 189year.</p></li>
 190
 191<li><p>Give a URL to the relevant standards document, where applicable.
 192(Sample links to SUSv4 and LSB are provided, feel free to link to other
 193documentation or standards as appropriate.)</p></li>
 194
 195<li><p>Update the USE_YOURCOMMAND(NEWTOY(yourcommand,"blah",0)) line.
 196The NEWTOY macro fills out this command's <a href="#toy_list">toy_list</a>
 197structure.  The arguments to the NEWTOY macro are:</p>
 198
 199<ol>
 200<li><p>the name used to run your command</p></li>
 201<li><p>the command line argument <a href="#lib_args">option parsing string</a> (0 if none)</p></li>
 202<li><p>a bitfield of TOYFLAG values
 203(defined in toys.h) providing additional information such as where your
 204command should be installed on a running system, whether to blank umask
 205before running, whether or not the command must run as root (and thus should
 206retain root access if installed SUID), and so on.</p></li>
 207</ol>
 208</li>
 209
 210<li><p>Change the kconfig data (from "config YOURCOMMAND" to the end of the
 211comment block) to supply your command's configuration and help
 212information.  The uppper case config symbols are used by menuconfig, and are
 213also what the CFG_ and USE_() macros are generated from (see [TODO]).  The
 214help information here is used by menuconfig, and also by the "help" command to
 215describe your new command.  (See [TODO] for details.)  By convention,
 216unfinished commands default to "n" and finished commands default to "y",
 217so "make defconfig" selects all finished commands.  (Note, "finished" means
 218"ready to be used", not that it'll never change again.)<p>
 219
 220<p>Each help block should start with a "usage: yourcommand" line explaining
 221any command line arguments added by this config option.  The "help" command
 222outputs this text, and scripts/config2help.c in the build infrastructure
 223collates these usage lines for commands with multiple configuration
 224options when producing generated/help.h.</p>
 225</li>
 226
 227<li><p>Change the "#define FOR_hello" line to "#define FOR_yourcommand" right
 228before the "#include <toys.h>". (This selects the appropriate FLAG_ macros and
 229does a "#define TT this.yourcommand" so you can access the global variables
 230out of the space-saving union of structures. If you aren't using any command
 231flag bits and aren't defining a GLOBAL block, you can delete this line.)</p></li>
 232
 233<li><p>Update the GLOBALS() macro to contain your command's global
 234variables. If your command has no global variables, delete this macro.</p>
 235
 236<p>Variables in the GLOBALS() block are are stored in a space saving
 237<a href="#toy_union">union of structures</a> format, which may be accessed
 238using the TT macro as if TT were a global structure (so TT.membername).
 239If you specified two-character command line arguments in
 240NEWTOY(), the first few global variables will be initialized by the automatic
 241argument parsing logic, and the type and order of these variables must
 242correspond to the arguments specified in NEWTOY().
 243(See <a href="#lib_args">lib/args.c</a> for details.)</p></li>
 244
 245<li><p>Rename hello_main() to yourcommand_main().  This is the main() function
 246where execution of your command starts. Your command line options are
 247already sorted into this.optflags, this.optargs, this.optc, and the GLOBALS()
 248as appropriate by the time this function is called. (See
 249<a href="#lib_args">get_optflags()</a> for details.)</p></li>
 250
 251<li><p>Switch on TOYBOX_DEBUG in menuconfig (toybox global settings menu)
 252the first time you build and run your new command. If anything is wrong
 253with your option string, that will give you error messages.</p>
 254
 255<p>Otherwise it'll just segfault without
 256explanation when it falls off the end because it didn't find a matching
 257end parantheses for a longopt, or you put a nonexistent option in a square
 258bracket grouping... Since these kind of errors can only be caused by a
 259developer, not by end users, we don't normally want runtime checks for
 260them. Once you're happy with your option string, you can switch TOYBOX_DEBUG
 261back off.</p></li>
 262</ul>
 263
 264<a name="headers" /><h2><a href="#headers">Headers.</a></h2>
 265
 266<p>Commands generally don't have their own headers. If it's common code
 267it can live in lib/, if it isn't put it in the command's .c file. (The line
 268between implementing multiple commands in a C file via OLDTOY() to share
 269infrastructure and moving that shared infrastructure to lib/ is a judgement
 270call. Try to figure out which is simplest.)</p>
 271
 272<p>The top level toys.h should #include all the standard (posix) headers
 273that any command uses. (Partly this is friendly to ccache and partly this
 274makes the command implementations shorter.) Individual commands should only
 275need to include nonstandard headers that might prevent that command from
 276building in some context we'd care about (and thus requiring that command to
 277be disabled to avoid a build break).</p>
 278
 279<p>Target-specific stuff (differences between compiler versions, libc versions,
 280or operating systems) should be confined to lib/portability.h and
 281lib/portability.c. (There's even some minimal compile-time environment probing
 282that writes data to generated/portability.h, see scripts/genconfig.sh.)</p>
 283
 284<p>Only include linux/*.h headers from individual commands (not from other
 285headers), and only if you really need to. Data that varies per architecture
 286is a good reason to include a header. If you just need a couple constants
 287that haven't changed since the 1990's, it's ok to #define them yourself or
 288just use the constant inline with a comment explaining what it is. (A
 289#define that's only used once isn't really helping.)</p>
 290
 291<p><a name="top" /><h1><a href="#top">Top level directory.</a></h1></p>
 292
 293<p>This directory contains global infrastructure.</p>
 294
 295<h3>toys.h</h3>
 296<p>Each command #includes "toys.h" as part of its standard prolog. It
 297may "#define FOR_commandname" before doing so to get some extra entries
 298specific to this command.</p>
 299
 300<p>This file sucks in most of the commonly used standard #includes, so
 301individual files can just #include "toys.h" and not have to worry about
 302stdargs.h and so on.  Individual commands still need to #include
 303special-purpose headers that may not be present on all systems (and thus would
 304prevent toybox from building that command on such a system with that command
 305enabled).  Examples include regex support, any "linux/" or "asm/" headers, mtab
 306support (mntent.h and sys/mount.h), and so on.</p>
 307
 308<p>The toys.h header also defines structures for most of the global variables
 309provided to each command by toybox_main().  These are described in
 310detail in the description for main.c, where they are initialized.</p>
 311
 312<p>The global variables are grouped into structures (and a union) for space
 313savings, to more easily track the amount of memory consumed by them,
 314so that they may be automatically cleared/initialized as needed, and so
 315that access to global variables is more easily distinguished from access to
 316local variables.</p>
 317
 318<h3>main.c</h3>
 319<p>Contains the main() function where execution starts, plus
 320common infrastructure to initialize global variables and select which command
 321to run.  The "toybox" multiplexer command also lives here.  (This is the
 322only command defined outside of the toys directory.)</p>
 323
 324<p>Execution starts in main() which trims any path off of the first command
 325name and calls toybox_main(), which calls toy_exec(), which calls toy_find()
 326and toy_init() before calling the appropriate command's function from
 327toy_list[] (via toys.which->toy_main()).
 328If the command is "toybox", execution recurses into toybox_main(), otherwise
 329the call goes to the appropriate commandname_main() from a C file in the toys
 330directory.</p>
 331
 332<p>The following global variables are defined in main.c:</p>
 333<ul>
 334<a name="toy_list" />
 335<li><p><b>struct toy_list toy_list[]</b> - array describing all the
 336commands currently configured into toybox.  The first entry (toy_list[0]) is
 337for the "toybox" multiplexer command, which runs all the other built-in commands
 338without symlinks by using its first argument as the name of the command to
 339run and the rest as that command's argument list (ala "./toybox echo hello").
 340The remaining entries are the commands in alphabetical order (for efficient
 341binary search).</p>
 342
 343<p>This is a read-only array initialized at compile time by
 344defining macros and #including generated/newtoys.h.</p>
 345
 346<p>Members of struct toy_list (defined in "toys.h") include:</p>
 347<ul>
 348<li><p>char *<b>name</b> - the name of this command.</p></li>
 349<li><p>void (*<b>toy_main</b>)(void) - function pointer to run this
 350command.</p></li>
 351<li><p>char *<b>options</b> - command line option string (used by
 352get_optflags() in lib/args.c to intialize toys.optflags, toys.optargs, and
 353entries in the toy's GLOBALS struct).  When this is NULL, no option
 354parsing is done before calling toy_main().</p></li>
 355<li><p>int <b>flags</b> - Behavior flags for this command.  The following flags are currently understood:</p>
 356
 357<ul>
 358<li><b>TOYFLAG_USR</b> - Install this command under /usr</li>
 359<li><b>TOYFLAG_BIN</b> - Install this command under /bin</li>
 360<li><b>TOYFLAG_SBIN</b> - Install this command under /sbin</li>
 361<li><b>TOYFLAG_NOFORK</b> - This command can be used as a shell builtin.</li>
 362<li><b>TOYFLAG_UMASK</b> - Call umask(0) before running this command.</li>
 363<li><b>TOYFLAG_STAYROOT</b> - Don't drop permissions for this command if toybox is installed SUID root.</li>
 364<li><b>TOYFLAG_NEEDROOT</b> - This command cannot function unless run with root access.</li>
 365</ul>
 366<br>
 367
 368<p>These flags are combined with | (or).  For example, to install a command
 369in /usr/bin, or together TOYFLAG_USR|TOYFLAG_BIN.</p>
 370</ul>
 371</li>
 372
 373<li><p><b>struct toy_context toys</b> - global structure containing information
 374common to all commands, initializd by toy_init() and defined in "toys.h".
 375Members of this structure include:</p>
 376<ul>
 377<li><p>struct toy_list *<b>which</b> - a pointer to this command's toy_list
 378structure.  Mostly used to grab the name of the running command
 379(toys->which.name).</p>
 380</li>
 381<li><p>int <b>exitval</b> - Exit value of this command.  Defaults to zero.  The
 382error_exit() functions will return 1 if this is zero, otherwise they'll
 383return this value.</p></li>
 384<li><p>char **<b>argv</b> - "raw" command line options, I.E. the original
 385unmodified string array passed in to main().  Note that modifying this changes
 386"ps" output, and is not recommended.  This array is null terminated; a NULL
 387entry indicates the end of the array.</p>
 388<p>Most commands don't use this field, instead the use optargs, optflags,
 389and the fields in the GLOBALS struct initialized by get_optflags().</p>
 390</li>
 391<li><p>unsigned <b>optflags</b> - Command line option flags, set by
 392<a href="#lib_args">get_optflags()</a>.  Indicates which of the command line options listed in
 393toys->which.options occurred this time.</p>
 394
 395<p>The rightmost command line argument listed in toys->which.options sets bit
 3961, the next one sets bit 2, and so on.  This means the bits are set in the same
 397order the binary digits would be listed if typed out as a string.  For example,
 398the option string "abcd" would parse the command line "-c" to set optflags to 2,
 399"-a" would set optflags to 8, and "-bd" would set optflags to 6 (4|2).</p>
 400
 401<p>Only letters are relevant to optflags.  In the string "a*b:c#d", d=1, c=2,
 402b=4, a=8.  Punctuation after a letter initializes global variables at the
 403start of the GLOBALS() block (see <a href="#toy_union">union toy_union this</a>
 404for details).</p>
 405
 406<p>The build infrastructure creates FLAG_ macros for each option letter,
 407corresponding to the bit position, so you can check (toys.optflags & FLAG_x)
 408to see if a flag was specified. (The correct set of FLAG_ macros is selected
 409by defining FOR_mycommand before #including toys.h. The macros live in
 410toys/globals.h which is generated by scripts/make.sh.)</p>
 411
 412<p>For more information on option parsing, see <a href="#lib_args">get_optflags()</a>.</p>
 413
 414</li>
 415<li><p>char **<b>optargs</b> - Null terminated array of arguments left over
 416after get_optflags() removed all the ones it understood.  Note: optarg[0] is
 417the first argument, not the command name.  Use toys.which->name for the command
 418name.</p></li>
 419<li><p>int <b>optc</b> - Optarg count, equivalent to argc but for
 420optargs[].<p></li>
 421</ul>
 422
 423<a name="toy_union" />
 424<li><p><b>union toy_union this</b> - Union of structures containing each
 425command's global variables.</p>
 426
 427<p>Global variables are useful: they reduce the overhead of passing extra
 428command line arguments between functions, they conveniently start prezeroed to
 429save initialization costs, and the command line argument parsing infrastructure
 430can also initialize global variables with its results.</p>
 431
 432<p>But since each toybox process can only run one command at a time, allocating
 433space for global variables belonging to other commands you aren't currently
 434running would be wasteful.</p>
 435
 436<p>Toybox handles this by encapsulating each command's global variables in
 437a structure, and declaring a union of those structures with a single global
 438instance (called "this").  The GLOBALS() macro contains the global
 439variables that should go in the current command's global structure.  Each
 440variable can then be accessed as "this.commandname.varname".
 441If you #defined FOR_commandname before including toys.h, the macro TT is
 442#defined to this.commandname so the variable can then be accessed as
 443"TT.variable".  See toys/hello.c for an example.</p>
 444
 445<p>A command that needs global variables should declare a structure to
 446contain them all, and add that structure to this union.  A command should never
 447declare global variables outside of this, because such global variables would
 448allocate memory when running other commands that don't use those global
 449variables.</p>
 450
 451<p>The first few fields of this structure can be intialized by <a href="#lib_args">get_optargs()</a>,
 452as specified by the options field off this command's toy_list entry.  See
 453the get_optargs() description in lib/args.c for details.</p>
 454</li>
 455
 456<li><b>char toybuf[4096]</b> - a common scratch space buffer guaranteed
 457to start zeroed, so commands don't need to allocate/initialize their own.
 458Any command is free to use this, and it should never be directly referenced
 459by functions in lib/ (although commands are free to pass toybuf in to a
 460library function as an argument).</li>
 461
 462<li><b>char libbuf[4096]</b> - like toybuf, but for use by common code in
 463lib/*.c. Commands should never directly reference libbuf, and library
 464could should nnever directly reference toybuf.</li>
 465</ul>
 466
 467<p>The following functions are defined in main.c:</p>
 468<ul>
 469<li><p>struct toy_list *<b>toy_find</b>(char *name) - Return the toy_list
 470structure for this command name, or NULL if not found.</p></li>
 471<li><p>void <b>toy_init</b>(struct toy_list *which, char *argv[]) - fill out
 472the global toys structure, calling get_optargs() if necessary.</p></li>
 473<li><p>void <b>toy_exec</b>(char *argv[]) - Run a built-in command with
 474arguments.</p>
 475<p>Calls toy_find() on argv[0] (which must be just a command name
 476without path).  Returns if it can't find this command, otherwise calls
 477toy_init(), toys->which.toy_main(), and exit() instead of returning.</p>
 478
 479<p>Use the library function xexec() to fall back to external executables
 480in $PATH if toy_exec() can't find a built-in command.  Note that toy_exec()
 481does not strip paths before searching for a command, so "./command" will
 482never match an internal command.</li>
 483
 484<li><p>void <b>toybox_main</b>(void) - the main function for the multiplexer
 485command (I.E. "toybox").  Given a command name as its first argument, calls
 486toy_exec() on its arguments.  With no arguments, it lists available commands.
 487If the first argument starts with "-" it lists each command with its default
 488install path prepended.</p></li>
 489
 490</ul>
 491
 492<h3>Config.in</h3>
 493
 494<p>Top level configuration file in a stylized variant of
 495<a href=http://kernel.org/doc/Documentation/kbuild/kconfig-language.txt>kconfig</a> format.  Includes generated/Config.in.</p>
 496
 497<p>These files are directly used by "make menuconfig" to select which commands
 498to build into toybox (thus generating a .config file), and by
 499scripts/config2help.py to create generated/help.h.</p>
 500
 501<a name="generated" />
 502<h1><a href="#generated">Temporary files:</a></h1>
 503
 504<p>There is one temporary file in the top level source directory:</p>
 505<ul>
 506<li><p><b>.config</b> - Configuration file generated by kconfig, indicating
 507which commands (and options to commands) are currently enabled.  Used
 508to make generated/config.h and determine which toys/*/*.c files to build.</p>
 509
 510<p>You can create a human readable "miniconfig" version of this file using
 511<a href=http://landley.net/aboriginal/new_platform.html#miniconfig>these
 512instructions</a>.</p>
 513</li>
 514</ul>
 515
 516<p><h2>Directory generated/</h2></p>
 517
 518<p>The remaining temporary files live in the "generated/" directory,
 519which is for files generated at build time from other source files.</p>
 520
 521<ul>
 522<li><p><b>generated/Config.in</b> - Kconfig entries for each command, included
 523from the top level Config.in. The help text here is used to generate
 524help.h.</p>
 525
 526<p>Each command has a configuration entry with an upper case version of
 527the command name. Options to commands start with the command
 528name followed by an underscore and the option name. Global options are attached
 529to the "toybox" command, and thus use the prefix "TOYBOX_".  This organization
 530is used by scripts/cfg2files to select which toys/*/*.c files to compile for a
 531given .config.</p>
 532</li>
 533
 534<li><p><b>generated/config.h</b> - list of CFG_SYMBOL and USE_SYMBOL() macros,
 535generated from .config by a sed invocation in scripts/make.sh.</p>
 536
 537<p>CFG_SYMBOL is a comple time constant set to 1 for enabled symbols and 0 for
 538disabled symbols. This allows the use of normal if() statements to remove
 539code at compile time via the optimizer's dead code elimination (which removes
 540from the binary any code that cannot be reached). This saves space without
 541cluttering the code with #ifdefs or leading to configuration dependent build
 542breaks. (See the 1992 Usenix paper
 543<a href=http://doc.cat-v.org/henry_spencer/ifdef_considered_harmful.pdf>#ifdef
 544Considered Harmful</a> for more information.)</p>
 545
 546<p>When you can't entirely avoid an #ifdef, the USE_SYMBOL(code) macro
 547provides a less intrusive alternative, evaluating to the code in parentheses
 548when the symbol is enabled, and nothing when the symbol is disabled. This
 549is most commonly used around NEWTOY() declarations (so only the enabled
 550commands show up in toy_list), and in option strings. This can also be used
 551for things like varargs or structure members which can't always be
 552eliminated by a simple test on CFG_SYMBOL. Remember, unlike CFG_SYMBOL
 553this is really just a variant of #ifdef, and can still result in configuration
 554dependent build breaks. Use with caution.</p>
 555</li>
 556
 557<li><p><b>generated/flags.h</b> - FLAG_? macros indicating which command
 558line options were seen. The option parsing in lib/args.c sets bits in
 559toys.optflags, which can be tested by anding with the appropriate FLAG_
 560macro. (Bare longopts, which have no corresponding short option, will
 561have the longopt name after FLAG_. All others use the single letter short
 562option.)</p>
 563
 564<p>To get the appropriate macros for your command, #define FOR_commandname
 565before #including toys.h. To switch macro sets (because you have an OLDTOY()
 566with different options in the same .c file), #define CLEANUP_oldcommand
 567and also #define FOR_newcommand, then #include "generated/flags.h" to switch.
 568</p>
 569</li>
 570
 571<li><p><b>generated/globals.h</b> -
 572Declares structures to hold the contents of each command's GLOBALS(),
 573and combines them into "global_union this". (Yes, the name was
 574chosen to piss off C++ developers who think that C
 575is merely a subset of C++, not a language in its own right.)</p>
 576
 577<p>The union reuses the same memory for each command's global struct:
 578since only one command's globals are in use at any given time, collapsing
 579them together saves space. The headers #define TT to the appropriate
 580"this.commandname", so you can refer to the current command's global
 581variables out of "this" as TT.variablename.</p>
 582
 583<p>The globals start zeroed, and the first few are filled out by the 
 584lib/args.c argument parsing code called from main.c.</p>
 585</li>
 586
 587<li><p><b>toys/help.h</b> - Help strings for use by the "help" command and
 588--help options. This file #defines a help_symbolname string for each
 589symbolname, but only the symbolnames matching command names get used
 590by show_help() in lib/help.c to display help for commands.</p>
 591
 592<p>This file is created by scripts/make.sh, which compiles scripts/config2help.c
 593into the binary generated/config2help, and then runs it against the top
 594level .config and Config.in files to extract the help text from each config
 595entry and collate together dependent options.</p>
 596
 597<p>This file contains help text for all commands, regardless of current
 598configuration, but only the ones currently enabled in the .config file
 599wind up in the help_data[] array, and only the enabled dependent options
 600have their help text added to the command they depend on.</p>
 601</li>
 602
 603<li><p><b>generated/newtoys.h</b> - 
 604All the NEWTOY() and OLDTOY() macros from toys/*/*.c. The "toybox" multiplexer
 605is the first entry, the rest are in alphabetical order. Each line should be
 606inside an appropriate USE_ macro, so code that #includes this file only sees
 607the currently enabled commands.</p>
 608
 609<p>By #definining NEWTOY() to various things before #including this file,
 610it may be used to create function prototypes (in toys.h), initialize the
 611help_data array (in lib/help.c),  initialize the toy_list array (in main.c,
 612the alphabetical order lets toy_find() do a binary search, the exception to
 613the alphabetical order lets it use the multiplexer without searching), and so
 614on.  (It's even used to initialize the NEED_OPTIONS macro, which produces a 1
 615or 0 for each command using command line option parsing, which is ORed together
 616to allow compile-time dead code elimination to remove the whole of
 617lib/args.c if nothing currently enabled is using it.)<p>
 618
 619<p>Each NEWTOY and OLDTOY macro contains the command name, command line
 620option string (telling lib/args.c how to parse command line options for
 621this command), recommended install location, and miscelaneous data such
 622as whether this command should retain root permissions if installed suid.</p>
 623</li>
 624
 625<li><p><b>toys/oldtoys.h</b> - Macros with the command line option parsing
 626string for each NEWTOY. This allows an OLDTOY that's just an alias for an
 627existing command to refer to the existing option string instead of
 628having to repeat it.</p>
 629</li>
 630</ul>
 631
 632<a name="lib">
 633<h2>Directory lib/</h2>
 634
 635<p>TODO: document lots more here.</p>
 636
 637<p>lib: getmountlist(), error_msg/error_exit, xmalloc(),
 638strlcpy(), xexec(), xopen()/xread(), xgetcwd(), xabspath(), find_in_path(),
 639itoa().</p>
 640
 641
 642
 643<a name="lib_xwrap"><h3>lib/xwrap.c</h3>
 644
 645<p>Functions prefixed with the letter x call perror_exit() when they hit
 646errors, to eliminate common error checking. This prints an error message
 647and the strerror() string for the errno encountered.</p>
 648
 649<p>We replaced exit(), _exit(), and atexit() with xexit(), _xexit(), and
 650sigatexit(). This gives _xexit() the option to siglongjmp(toys.rebound, 1)
 651instead of exiting, lets xexit() report stdout flush failures to stderr
 652and change the exit code to indicate error, lets our toys.exit function
 653change happen for signal exit paths and lets us remove the functions
 654after we've called them.</p>
 655
 656<p>You can intercept our exit by assigning a setjmp/longjmp buffer to
 657toys.rebound (set it back to zero to restore the default behavior).
 658If you do this, cleaning up resource leaks is your problem.</p>
 659
 660<ul>
 661<li><b>void xstrncpy(char *dest, char *src, size_t size)</b></li>
 662<li><p><b><p>void _xexit(void)</b></p>
 663<p>Calls siglongjmp(toys.rebound, 1), or else _exit(toys.exitval). This
 664lets you ignore errors with the NO_EXIT() macro wrapper, or intercept
 665them with WOULD_EXIT().</p>
 666<li><b><p>void xexit(void)</b></p>
 667<p>Calls toys.xexit functions (if any) and flushes stdout/stderr (reporting
 668failure to write to stdout both to stderr and in the exit code), then
 669calls _xexit().</p>
 670</li>
 671<li><b>void *xmalloc(size_t size)</b></li>
 672<li><b>void *xzalloc(size_t size)</b></li>
 673<li><b>void *xrealloc(void *ptr, size_t size)</b></li>
 674<li><b>char *xstrndup(char *s, size_t n)</b></li>
 675<li><b>char *xstrdup(char *s)</b></li>
 676<li><b>char *xmprintf(char *format, ...)</b></li>
 677<li><b>void xprintf(char *format, ...)</b></li>
 678<li><b>void xputs(char *s)</b></li>
 679<li><b>void xputc(char c)</b></li>
 680<li><b>void xflush(void)</b></li>
 681<li><b>pid_t xfork(void)</b></li>
 682<li><b>void xexec_optargs(int skip)</b></li>
 683<li><b>void xexec(char **argv)</b></li>
 684<li><b>pid_t xpopen(char **argv, int *pipes)</b></li>
 685<li><b>int xpclose(pid_t pid, int *pipes)</b></li>
 686<li><b>void xaccess(char *path, int flags)</b></li>
 687<li><b>void xunlink(char *path)</b></li>
 688<li><p><b>int xcreate(char *path, int flags, int mode)<br />
 689int xopen(char *path, int flags)</b></p>
 690
 691<p>The xopen() and xcreate() functions open an existing file (exiting if
 692it's not there) and create a new file (exiting if it can't).</p>
 693
 694<p>They default to O_CLOEXEC so the filehandles aren't passed on to child
 695processes. Feed in O_CLOEXEC to disable this.</p>
 696</li>
 697<li><p><b>void xclose(int fd)</b></p>
 698
 699<p>Because NFS is broken, and won't necessarily perform the requested
 700operation (and report the error) until you close the file. Of course, this
 701being NFS, it's not guaranteed to report the error there either, but it
 702_can_.</p>
 703
 704<p>Nothing else ever reports an error on close, everywhere else it's just a
 705VFS operation freeing some resources. NFS is _special_, in a way that
 706other network filesystems like smbfs and v9fs aren't..</p>
 707</li>
 708<li><b>int xdup(int fd)</b></li>
 709<li><p><b>size_t xread(int fd, void *buf, size_t len)</b></p>
 710
 711<p>Can return 0, but not -1.</p>
 712</li>
 713<li><p><b>void xreadall(int fd, void *buf, size_t len)</b></p>
 714
 715<p>Reads the entire len-sized buffer, retrying to complete short
 716reads. Exits if it can't get enough data.</p></li>
 717
 718<li><p><b>void xwrite(int fd, void *buf, size_t len)</b></p>
 719
 720<p>Retries short writes, exits if can't write the entire buffer.</p></li>
 721
 722<li><b>off_t xlseek(int fd, off_t offset, int whence)</b></li>
 723<li><b>char *xgetcwd(void)</b></li>
 724<li><b>void xstat(char *path, struct stat *st)</b></li>
 725<li><p><b>char *xabspath(char *path, int exact) </b></p>
 726
 727<p>After several years of
 728<a href=http://landley.net/notes-2007.html#18-06-2007>wrestling</a>
 729<a href=http://landley.net/notes-2008.html#19-01-2008>with</a> realpath(), 
 730I broke down and <a href=http://landley.net/notes-2012.html#20-11-2012>wrote
 731my own</a> implementation that doesn't use the one in libc. As I explained:
 732
 733<blockquote><p>If the path ends with a broken link,
 734readlink -f should show where the link points to, not where the broken link
 735lives. (The point of readlink -f is "if I write here, where would it attempt
 736to create a file".) The problem is, realpath() returns NULL for a path ending
 737with a broken link, and I can't beat different behavior out of code locked
 738away in libc.</p></blockquote>
 739
 740<p>
 741</li>
 742<li><b>void xchdir(char *path)</b></li>
 743<li><b>void xchroot(char *path)</b></li>
 744
 745<li><p><b>struct passwd *xgetpwuid(uid_t uid)<br />
 746struct group *xgetgrgid(gid_t gid)<br />
 747struct passwd *xgetpwnam(char *name)</b></p>
 748</li>
 749
 750<li><b>void xsetuser(struct passwd *pwd)</b></li>
 751<li><b>char *xreadlink(char *name)</b></li>
 752<li><b>char *xreadfile(char *name, char *buf, off_t len)</b></li>
 753<li><b>int xioctl(int fd, int request, void *data)</b></li>
 754<li><b>void xpidfile(char *name)</b></li>
 755<li><b>void xsendfile(int in, int out)</b></li>
 756<li><b>long xparsetime(char *arg, long units, long *fraction)</b></li>
 757<li><b>void xregcomp(regex_t *preg, char *regex, int cflags)</b></li>
 758</ul>
 759
 760<a name="lib_lib"><h3>lib/lib.c</h3>
 761<p>Eight gazillion common functions, see lib/lib.h for the moment:</p>
 762
 763<h3>lib/portability.h</h3>
 764
 765<p>This file is automatically included from the top of toys.h, and smooths
 766over differences between platforms (hardware targets, compilers, C libraries,
 767operating systems, etc).</p>
 768
 769<p>This file provides SWAP macros (SWAP_BE16(x) and SWAP_LE32(x) and so on).</p>
 770
 771<p>A macro like SWAP_LE32(x) means "The value in x is stored as a little
 772endian 32 bit value, so perform the translation to/from whatever the native
 77332-bit format is".  You do the swap once on the way in, and once on the way
 774out. If your target is already little endian, the macro is a NOP.</p>
 775
 776<p>The SWAP macros come in BE and LE each with 16, 32, and 64 bit versions.
 777In each case, the name of the macro refers to the _external_ representation,
 778and converts to/from whatever your native representation happens to be (which
 779can vary depending on what you're currently compiling for).</p>
 780
 781<a name="lib_llist"><h3>lib/llist.c</h3>
 782
 783<p>Some generic single and doubly linked list functions, which take
 784advantage of a couple properties of C:</p>
 785
 786<ul>
 787<li><p>Structure elements are laid out in memory in the order listed, and
 788the first element has no padding. This means you can always treat (typecast)
 789a pointer to a structure as a pointer to the first element of the structure,
 790even if you don't know anything about the data following it.</p></li>
 791
 792<li><p>An array of length zero at the end of a structure adds no space
 793to the sizeof() the structure, but if you calculate how much extra space
 794you want when you malloc() the structure it will be available at the end.
 795Since C has no bounds checking, this means each struct can have one variable
 796length array.</p></li>
 797</ul>
 798
 799<p>Toybox's list structures always have their <b>next</b> pointer as
 800the first entry of each struct, and singly linked lists end with a NULL pointer.
 801This allows generic code to traverse such lists without knowing anything
 802else about the specific structs composing them: if your pointer isn't NULL
 803typecast it to void ** and dereference once to get the next entry.</p>
 804
 805<p><b>lib/lib.h</b> defines three structure types:</p>
 806<ul>
 807<li><p><b>struct string_list</b> - stores a single string (<b>char str[0]</b>),
 808memory for which is allocated as part of the node. (I.E. llist_traverse(list,
 809free); can clean up after this type of list.)</p></li>
 810
 811<li><p><b>struct arg_list</b> - stores a pointer to a single string
 812(<b>char *arg</b>) which is stored in a separate chunk of memory.</p></li>
 813
 814<li><p><b>struct double_list</b> - has a second pointer (<b>struct double_list
 815*prev</b> along with a <b>char *data</b> for payload.</p></li>
 816</ul>
 817
 818<b>List Functions</b>
 819
 820<ul>
 821<li><p>void *<b>llist_pop</b>(void **list) - advances through a list ala
 822<b>node = llist_pop(&list);</b>  This doesn't modify the list contents,
 823but does advance the pointer you feed it (which is why you pass the _address_
 824of that pointer, not the pointer itself).</p></li>
 825
 826<li><p>void <b>llist_traverse</b>(void *list, void (*using)(void *data)) -
 827iterate through a list calling a function on each node.</p></li>
 828
 829<li><p>struct double_list *<b>dlist_add</b>(struct double_list **llist, char *data)
 830- append an entry to a circular linked list.
 831This function allocates a new struct double_list wrapper and returns the
 832pointer to the new entry (which you can usually ignore since it's llist->prev,
 833but if llist was NULL you need it). The argument is the ->data field for the
 834new node.</p></li>
 835<ul><li><p>void <b>dlist_add_nomalloc</b>(struct double_list **llist,
 836struct double_list *new) - append existing struct double_list to
 837list, does not allocate anything.</p></li></ul>
 838</ul>
 839
 840<b>List code trivia questions:</b>
 841
 842<ul>
 843<li><p><b>Why do arg_list and double_list contain a char * payload instead of
 844a void *?</b> - Because you always have to typecast a void * to use it, and
 845typecasting a char * does no harm. Since strings are the most common
 846payload, and doing math on the pointer ala
 847"(type *)(ptr+sizeof(thing)+sizeof(otherthing))" requires ptr to be char *
 848anyway (at least according to the C standard), defaulting to char * saves
 849a typecast.</p>
 850</li>
 851
 852<li><p><b>Why do the names ->str, ->arg, and ->data differ?</b> - To force
 853you to keep track of which one you're using, calling free(node->str) would
 854be bad, and _failing_ to free(node->arg) leaks memory.</p></li>
 855
 856<li><p><b>Why does llist_pop() take a void * instead of void **?</b> -
 857because the stupid compiler complains about "type punned pointers" when
 858you typecast and dereference on the same line,
 859due to insane FSF developers hardwiring limitations of their optimizer
 860into gcc's warning system. Since C automatically typecasts any other
 861pointer type to and from void *, the current code works fine. It's sad that it
 862won't warn you if you forget the &, but the code crashes pretty quickly in
 863that case.</p></li>
 864
 865<li><p><b>How do I assemble a singly-linked-list in order?</b> - use
 866a double_list, dlist_add() your entries, and then call dlist_terminate(list)
 867to break the circle when done (turning the last ->next and the first ->prev
 868into NULLs).</p>
 869</ul>
 870
 871<a name="lib_args"><h3>lib/args.c</h3>
 872
 873<p>Toybox's main.c automatically parses command line options before calling the
 874command's main function. Option parsing starts in get_optflags(), which stores
 875results in the global structures "toys" (optflags and optargs) and "this".</p>
 876
 877<p>The option parsing infrastructure stores a bitfield in toys.optflags to
 878indicate which options the current command line contained, and defines FLAG
 879macros code can use to check whether each argument's bit is set. Arguments
 880attached to those options are saved into the command's global structure
 881("this"). Any remaining command line arguments are collected together into
 882the null-terminated array toys.optargs, with the length in toys.optc. (Note
 883that toys.optargs does not contain the current command name at position zero,
 884use "toys.which->name" for that.) The raw command line arguments get_optflags()
 885parsed are retained unmodified in toys.argv[].</p>
 886
 887<p>Toybox's option parsing logic is controlled by an "optflags" string, using
 888a format reminiscent of getopt's optargs but with several important differences.
 889Toybox does not use the getopt()
 890function out of the C library, get_optflags() is an independent implementation
 891which doesn't permute the original arguments (and thus doesn't change how the
 892command is displayed in ps and top), and has many features not present in
 893libc optargs() (such as the ability to describe long options in the same string
 894as normal options).</p>
 895
 896<p>Each command's NEWTOY() macro has an optflags string as its middle argument,
 897which sets toy_list.options for that command to tell get_optflags() what
 898command line arguments to look for, and what to do with them.
 899If a command has no option
 900definition string (I.E. the argument is NULL), option parsing is skipped
 901for that command, which must look at the raw data in toys.argv to parse its
 902own arguments. (If no currently enabled command uses option parsing,
 903get_optflags() is optimized out of the resulting binary by the compiler's
 904--gc-sections option.)</p>
 905
 906<p>You don't have to free the option strings, which point into the environment
 907space (I.E. the string data is not copied). A TOYFLAG_NOFORK command
 908that uses the linked list type "*" should free the list objects but not
 909the data they point to, via "llist_free(TT.mylist, NULL);". (If it's not
 910NOFORK, exit() will free all the malloced data anyway unless you want
 911to implement a CONFIG_TOYBOX_FREE cleanup for it.)</p>
 912
 913<h4>Optflags format string</h4>
 914
 915<p>Note: the optflags option description string format is much more
 916concisely described by a large comment at the top of lib/args.c.</p>
 917
 918<p>The general theory is that letters set optflags, and punctuation describes
 919other actions the option parsing logic should take.</p>
 920
 921<p>For example, suppose the command line <b>command -b fruit -d walrus -a 42</b>
 922is parsed using the optflags string "<b>a#b:c:d</b>".  (I.E.
 923toys.which->options="a#b:c:d" and argv = ["command", "-b", "fruit", "-d",
 924"walrus", "-a", "42"]).  When get_optflags() returns, the following data is
 925available to command_main():
 926
 927<ul>
 928<li><p>In <b>struct toys</b>:
 929<ul>
 930<li>toys.optflags = 13; // FLAG_a = 8 | FLAG_b = 4 | FLAG_d = 1</li>
 931<li>toys.optargs[0] = "walrus"; // leftover argument</li>
 932<li>toys.optargs[1] = NULL; // end of list</li>
 933<li>toys.optc = 1; // there was 1 leftover argument</li>
 934<li>toys.argv[] = {"-b", "fruit", "-d", "walrus", "-a", "42"}; // The original command line arguments
 935</ul>
 936<p></li>
 937
 938<li><p>In <b>union this</b> (treated as <b>long this[]</b>):
 939<ul>
 940<li>this[0] = NULL; // -c didn't get an argument this time, so get_optflags() didn't change it and toys_init() zeroed "this" during setup.)</li>
 941<li>this[1] = (long)"fruit"; // argument to -b</li>
 942<li>this[2] = 42; // argument to -a</li>
 943</ul>
 944</p></li>
 945</ul>
 946
 947<p>If the command's globals are:</p>
 948
 949<blockquote><pre>
 950GLOBALS(
 951        char *c;
 952        char *b;
 953        long a;
 954)
 955</pre></blockquote>
 956
 957<p>That would mean TT.c == NULL, TT.b == "fruit", and TT.a == 42.  (Remember,
 958each entry that receives an argument must be a long or pointer, to line up
 959with the array position.  Right to left in the optflags string corresponds to
 960top to bottom in GLOBALS().</p>
 961
 962<p>Put globals not filled out by the option parsing logic at the end of the
 963GLOBALS block. Common practice is to list the options one per line (to
 964make the ordering explicit, first to last in globals corresponds to right
 965to left in the option string), then leave a blank line before any non-option
 966globals.</p>
 967
 968<p><b>long toys.optflags</b></p>
 969
 970<p>Each option in the optflags string corresponds to a bit position in
 971toys.optflags, with the same value as a corresponding binary digit.  The
 972rightmost argument is (1<<0), the next to last is (1<<1) and so on.  If
 973the option isn't encountered while parsing argv[], its bit remains 0.</p>
 974
 975<p>Each option -x has a FLAG_x macro for the command letter. Bare --longopts
 976with no corresponding short option have a FLAG_longopt macro for the long
 977optionname. Commands enable these macros by #defining FOR_commandname before
 978#including <toys.h>. When multiple commands are implemented in the same
 979source file, you can switch flag contexts later in the file by
 980#defining CLEANUP_oldcommand and #defining FOR_newcommand, then
 981#including <generated/flags.h>.</p>
 982
 983<p>Options disabled in the current configuration (wrapped in
 984a USE_BLAH() macro for a CONFIG_BLAH that's switched off) have their
 985corresponding FLAG macro set to zero, so code checking them ala
 986if (toys.optargs & FLAG_x) gets optimized out via dead code elimination.
 987#defining FORCE_FLAGS when switching flag context disables this
 988behavior: the flag is never zero even if the config is disabled. This
 989allows code shared between multiple commands to use the same flag
 990values, as long as the common flags match up right to left in both option
 991strings.</p>
 992
 993<p>For example,
 994the optflags string "abcd" would parse the command line argument "-c" to set
 995optflags to 2, "-a" would set optflags to 8, "-bd" would set optflags to
 9966 (I.E. 4|2), and "-a -c" would set optflags to 10 (2|8). To check if -c
 997was encountered, code could test: if (toys.optflags & FLAG_c) printf("yup");
 998(See the toys/examples directory for more.)</p>
 999
1000<p>Only letters are relevant to optflags, punctuation is skipped: in the
1001string "a*b:c#d", d=1, c=2, b=4, a=8. The punctuation after a letter
1002usually indicate that the option takes an argument.</p>
1003
1004<p>Since toys.optflags is an unsigned int, it only stores 32 bits. (Which is
1005the amount a long would have on 32-bit platforms anyway; 64 bit code on
100632 bit platforms is too expensive to require in common code used by almost
1007all commands.) Bit positions beyond the 1<<31 aren't recorded, but
1008parsing higher options can still set global variables.</p>
1009
1010<p><b>Automatically setting global variables from arguments (union this)</b></p>
1011
1012<p>The following punctuation characters may be appended to an optflags
1013argument letter, indicating the option takes an additional argument:</p>
1014
1015<ul>
1016<li><b>:</b> - plus a string argument, keep most recent if more than one.</li>
1017<li><b>*</b> - plus a string argument, appended to a linked list.</li>
1018<li><b>@</b> - plus an occurrence counter (stored in a long)</li>
1019<li><b>#</b> - plus a signed long argument.
1020<li><b>-</b> - plus a signed long argument defaulting to negative (start argument with + to force a positive value).</li>
1021<li><b>.</b> - plus a floating point argument (if CFG_TOYBOX_FLOAT).</li>
1022<ul>The following can be appended to a float or double:
1023<li><b>&lt;123</b> - error if argument is less than this</li>
1024<li><b>&gt;123</b> - error if argument is greater than this</li>
1025<li><b>=123</b> - default value if argument not supplied</li>
1026</ul>
1027</ul>
1028
1029<p><b>GLOBALS</b></p>
1030
1031<p>Options which have an argument fill in the corresponding slot in the global
1032union "this" (see generated/globals.h), treating it as an array of longs
1033with the rightmost saved in this[0].  As described above, using "a*b:c#d",
1034"-c 42" would set this[0] = 42; and "-b 42" would set this[1] = "42"; each
1035slot is left NULL if the corresponding argument is not encountered.</p>
1036
1037<p>This behavior is useful because the LP64 standard ensures long and pointer
1038are the same size. C99 guarantees structure members will occur in memory
1039in the same order they're declared, and that padding won't be inserted between
1040consecutive variables of register size.  Thus the first few entries can
1041be longs or pointers corresponding to the saved arguments.</p>
1042
1043<p>The main downside is that numeric arguments ("#" and "-" format)
1044are limited to +- 2 billion on 32 bit platforms (the "truncate -s 8G"
1045problem), because long is only 64 bits on 64 bit hosts, so the capabilities
1046of some tools differ when built in 32 bit vs 64 bit mode. Fixing this
1047kind of ugly and even embedded designs are slowly moving to 64 bits,
1048so our current plan is to document the problem and wait it out. (If
1049"x32 mode" and similar becomes popular enough, we may revisit this
1050decision.)</p>
1051
1052<p>See toys/example/*.c for longer examples of parsing options into the
1053GLOBALS block.</p>
1054
1055<p><b>char *toys.optargs[]</b></p>
1056
1057<p>Command line arguments in argv[] which are not consumed by option parsing
1058(I.E. not recognized either as -flags or arguments to -flags) will be copied
1059to toys.optargs[], with the length of that array in toys.optc.
1060(When toys.optc is 0, no unrecognized command line arguments remain.)
1061The order of entries is preserved, and as with argv[] this new array is also
1062terminated by a NULL entry.</p>
1063
1064<p>Option parsing can require a minimum or maximum number of optargs left
1065over, by adding "<1" (read "at least one") or ">9" ("at most nine") to the
1066start of the optflags string.</p>
1067
1068<p>The special argument "--" terminates option parsing, storing all remaining
1069arguments in optargs.  The "--" itself is consumed.</p>
1070
1071<p><b>Other optflags control characters</b></p>
1072
1073<p>The following characters may occur at the start of each command's
1074optflags string, before any options that would set a bit in toys.optflags:</p>
1075
1076<ul>
1077<li><b>^</b> - stop at first nonoption argument (for nice, xargs...)</li>
1078<li><b>?</b> - allow unknown arguments (pass non-option arguments starting
1079with - through to optargs instead of erroring out).</li>
1080<li><b>&amp;</b> - the first argument has imaginary dash (ala tar/ps.  If given twice, all arguments have imaginary dash.)</li>
1081<li><b>&lt;</b> - must be followed by a decimal digit indicating at least this many leftover arguments are needed in optargs (default 0)</li>
1082<li><b>&gt;</b> - must be followed by a decimal digit indicating at most this many leftover arguments allowed (default MAX_INT)</li>
1083</ul>
1084
1085<p>The following characters may be appended to an option character, but do
1086not by themselves indicate an extra argument should be saved in this[].
1087(Technically any character not recognized as a control character sets an
1088optflag, but letters are never control characters.)</p>
1089
1090<ul>
1091<li><b>^</b> - stop parsing options after encountering this option, everything else goes into optargs.</li>
1092<li><b>|</b> - this option is required.  If more than one marked, only one is required.</li>
1093</ul>
1094
1095<p>The following may be appended to a float or double:</p>
1096
1097<ul>
1098<li><b>&lt;123</b> - error if argument is less than this</li>
1099<li><b>&gt;123</b> - error if argument is greater than this</li>
1100<li><b>=123</b> - default value if argument not supplied</li>
1101</ul>
1102
1103<p>Option parsing only understands <>= after . when CFG_TOYBOX_FLOAT
1104is enabled. (Otherwise the code to determine where floating point constants
1105end drops out.  When disabled, it can reserve a global data slot for the
1106argument so offsets won't change, but will never fill it out.) You can handle
1107this by using the USE_BLAH() macros with C string concatenation, ala:</p>
1108
1109<blockquote>"abc." USE_TOYBOX_FLOAT("<1.23>4.56=7.89") "def"</blockquote>
1110
1111<p><b>--longopts</b></p>
1112
1113<p>The optflags string can contain long options, which are enclosed in
1114parentheses. They may be appended to an existing option character, in
1115which case the --longopt is a synonym for that option, ala "a:(--fred)"
1116which understands "-a blah" or "--fred blah" as synonyms.</p>
1117
1118<p>Longopts may also appear before any other options in the optflags string,
1119in which case they have no corresponding short argument, but instead set
1120their own bit based on position. So for "(walrus)#(blah)xy:z", "command
1121--walrus 42" would set toys.optflags = 16 (-z = 1, -y = 2, -x = 4, --blah = 8)
1122and would assign this[1] = 42;</p>
1123
1124<p>A short option may have multiple longopt synonyms, "a(one)(two)", but
1125each "bare longopt" (ala "(one)(two)abc" before any option characters)
1126always sets its own bit (although you can group them with +X).</p>
1127
1128<p>Only bare longopts have a FLAG_ macro with the longopt name
1129(ala --fred would #define FLAG_fred). Other longopts use the short
1130option's FLAG macro to test the toys.optflags bit.</p>
1131
1132<p>Options with a semicolon ";" after their data type can only set their
1133corresponding GLOBALS() entry via "--longopt=value". For example, option
1134string "x(boing): y" would set TT.x if it saw "--boing=value", but would
1135treat "--boing value" as setting FLAG_x in toys.optargs, leaving TT.x NULL,
1136and keeping "value" in toys.optargs[]. (This lets "ls --color" and
1137"ls --color=auto" both work.)</p>
1138
1139<p><b>[groups]</b></p>
1140
1141<p>At the end of the option string, square bracket groups can define
1142relationships between existing options. (This only applies to short
1143options, bare --longopts can't participate.)</p>
1144
1145<p>The first character of the group defines the type, the remaining
1146characters are options it applies to:</p>
1147
1148<ul>
1149<li><b>-</b> - Exclusive, switch off all others in this group.</li>
1150<li><b>+</b> - Inclusive, switch on all others in this group.</li>
1151<li><b>!</b> - Error, fail if more than one defined.</li>
1152</ul>
1153
1154<p>So "abc[-abc]" means -ab = -b, -ba = -a, -abc = -c. "abc[+abc]"
1155means -ab=-abc, -c=-abc, and "abc[!abc] means -ab calls error_exit("no -b
1156with -a"). Note that [-] groups clear the GLOBALS option slot of
1157options they're switching back off, but [+] won't set options it didn't see
1158(just the optflags).</p>
1159
1160<p><b>whitespace</b></p>
1161
1162<p>Arguments may occur with or without a space (I.E. "-a 42" or "-a42").
1163The command line argument "-abc" may be interepreted many different ways:
1164the optflags string "cba" sets toys.optflags = 7, "c:ba" sets toys.optflags=4
1165and saves "ba" as the argument to -c, and "cb:a" sets optflags to 6 and saves
1166"c" as the argument to -b.</p>
1167
1168<p>Note that &amp; changes whitespace handling, so that the command line
1169"tar cvfCj outfile.tar.bz2 topdir filename" is parsed the same as
1170"tar filename -c -v -j -f outfile.tar.bz2 -C topdir". Note that "tar -cvfCj
1171one two three" would equal "tar -c -v -f Cj one two three". (This matches
1172historical usage.)</p>
1173
1174<p>Appending a space to the option in the option string ("a: b") makes it
1175require a space, I.E. "-ab" is interpreted as "-a" "-b". That way "kill -stop"
1176differs from "kill -s top".</p>
1177
1178<p>Appending ; to a longopt in the option string makes its argument optional,
1179and only settable with =, so in ls "(color):;" can accept "ls --color" and
1180"ls --color=auto" without complaining that the first has no argument.</p>
1181
1182<a name="lib_dirtree"><h3>lib/dirtree.c</h3>
1183
1184<p>The directory tree traversal code should be sufficiently generic
1185that commands never need to use readdir(), scandir(), or the fts.h family
1186of functions.</p>
1187
1188<p>These functions do not call chdir() or rely on PATH_MAX. Instead they
1189use openat() and friends, using one filehandle per directory level to
1190recurse into subdirectories. (I.E. they can descend 1000 directories deep
1191if setrlimit(RLIMIT_NOFILE) allows enough open filehandles, and the default
1192in /proc/self/limits is generally 1024.)</p>
1193
1194<p>There are two main ways to use dirtree: 1) assemble a tree of nodes
1195representing a snapshot of directory state and traverse them using the
1196->next and ->child pointers, or 2) traverse the tree calling a callback
1197function on each entry, and freeing its node afterwards. (You can also
1198combine the two, using the callback as a filter to determine which nodes
1199to keep.)</p>
1200
1201<p>The basic dirtree functions are:</p>
1202
1203<ul>
1204<li><p><b>struct dirtree *dirtree_read(char *path, int (*callback)(struct
1205dirtree node))</b> - recursively read files and directories, calling
1206callback() on each, and returning a tree of saved nodes (if any).
1207If path doesn't exist, returns DIRTREE_ABORTVAL. If callback is NULL,
1208returns a single node at that path.</p>
1209
1210<li><p><b>dirtree_notdotdot(struct dirtree *new)</b> - standard callback
1211which discards "." and ".." entries and returns DIRTREE_SAVE|DIRTREE_RECURSE
1212for everything else. Used directly, this assembles a snapshot tree of
1213the contents of this directory and its subdirectories
1214to be processed after dirtree_read() returns (by traversing the
1215struct dirtree's ->next and ->child pointers from the returned root node).</p>
1216
1217<li><p><b>dirtree_path(struct dirtree *node, int *plen)</b> - malloc() a
1218string containing the path from the root of this tree to this node. If
1219plen isn't NULL then *plen is how many extra bytes to malloc at the end
1220of string.</p></li>
1221
1222<li><p><b>dirtree_parentfd(struct dirtree *node)</b> - return fd of
1223directory containing this node, for use with openat() and such.</p></li>
1224</ul>
1225
1226<p>The <b>dirtree_read()</b> function is the standard way to start
1227directory traversal. It takes two arguments: a starting path for
1228the root of the tree, and a callback function. The callback() is called
1229on each directory entry, its argument is a fully populated
1230<b>struct dirtree *</b> (from lib/lib.h) describing the node, and its
1231return value tells the dirtree infrastructure what to do next.</p>
1232
1233<p>(There's also a three argument version,
1234<b>dirtree_flagread(char *path, int flags, int (*callback)(struct
1235dirtree node))</b>, which lets you apply flags like DIRTREE_SYMFOLLOW and
1236DIRTREE_SHUTUP to reading the top node, but this only affects the top node.
1237Child nodes use the flags returned by callback().</p>
1238
1239<p><b>struct dirtree</b></p>
1240
1241<p>Each struct dirtree node contains <b>char name[]</b> and <b>struct stat
1242st</b> entries describing a file, plus a <b>char *symlink</b>
1243which is NULL for non-symlinks.</p>
1244
1245<p>During a callback function, the <b>int dirfd</b> field of directory nodes
1246contains a directory file descriptor (for use with the openat() family of
1247functions). This isn't usually used directly, intstead call dirtree_parentfd()
1248on the callback's node argument. The <b>char again</b> field is 0 for the
1249first callback on a node, and 1 on the second callback (triggered by returning
1250DIRTREE_COMEAGAIN on a directory, made after all children have been processed).
1251</p>
1252
1253<p>Users of this code may put anything they like into the <b>long extra</b>
1254field. For example, "cp" and "mv" use this to store a dirfd for the destination
1255directory (and use DIRTREE_COMEAGAIN to get the second callback so they can
1256close(node->extra) to avoid running out of filehandles).
1257This field is not directly used by the dirtree code, and
1258thanks to LP64 it's large enough to store a typecast pointer to an
1259arbitrary struct.</p>
1260
1261<p>The return value of the callback combines flags (with boolean or) to tell
1262the traversal infrastructure how to behave:</p>
1263
1264<ul>
1265<li><p><b>DIRTREE_SAVE</b> - Save this node, assembling a tree. (Without
1266this the struct dirtree is freed after the callback returns. Filtering out
1267siblings is fine, but discarding a parent while keeping its child leaks
1268memory.)</p></li>
1269<li><p><b>DIRTREE_ABORT</b> - Do not examine any more entries in this
1270directory. (Does not propagate up tree: to abort entire traversal,
1271return DIRTREE_ABORT from parent callbacks too.)</p></li>
1272<li><p><b>DIRTREE_RECURSE</b> - Examine directory contents. Ignored for
1273non-directory entries. The remaining flags only take effect when
1274recursing into the children of a directory.</p></li>
1275<li><p><b>DIRTREE_COMEAGAIN</b> - Call the callback on this node a second time
1276after examining all directory contents, allowing depth-first traversal.
1277On the second call, dirtree->again is nonzero.</p></li>
1278<li><p><b>DIRTREE_SYMFOLLOW</b> - follow symlinks when populating children's
1279<b>struct stat st</b> (by feeding a nonzero value to the symfollow argument of
1280dirtree_add_node()), which means DIRTREE_RECURSE treats symlinks to
1281directories as directories. (Avoiding infinite recursion is the callback's
1282problem: the non-NULL dirtree->symlink can still distinguish between
1283them. The "find" command follows ->parent up the tree to the root node
1284each time, checking to make sure that stat's dev and inode pair don't
1285match any ancestors.)</p></li>
1286</ul>
1287
1288<p>Each struct dirtree contains three pointers (next, parent, and child)
1289to other struct dirtree.</p>
1290
1291<p>The <b>parent</b> pointer indicates the directory
1292containing this entry; even when not assembling a persistent tree of
1293nodes the parent entries remain live up to the root of the tree while
1294child nodes are active. At the top of the tree the parent pointer is
1295NULL, meaning the node's name[] is either an absolute path or relative
1296to cwd. The function dirtree_parentfd() gets the directory file descriptor
1297for use with openat() and friends, returning AT_FDCWD at the top of tree.</p>
1298
1299<p>The <b>child</b> pointer points to the first node of the list of contents of
1300this directory. If the directory contains no files, or the entry isn't
1301a directory, child is NULL.</p>
1302
1303<p>The <b>next</b> pointer indicates sibling nodes in the same directory as this
1304node, and since it's the first entry in the struct the llist.c traversal
1305mechanisms work to iterate over sibling nodes. Each dirtree node is a
1306single malloc() (even char *symlink points to memory at the end of the node),
1307so llist_free() works but its callback must descend into child nodes (freeing
1308a tree, not just a linked list), plus whatever the user stored in extra.</p>
1309
1310<p>The <b>dirtree_flagread</b>() function is a simple wrapper, calling <b>dirtree_add_node</b>()
1311to create a root node relative to the current directory, then calling
1312<b>dirtree_handle_callback</b>() on that node (which recurses as instructed by the callback
1313return flags). The flags argument primarily lets you
1314control whether or not to follow symlinks to the root node; symlinks
1315listed on the command line are often treated differently than symlinks
1316encountered during recursive directory traversal.
1317
1318<p>The ls command not only bypasses this wrapper, but never returns
1319<b>DIRTREE_RECURSE</b> from the callback, instead calling <b>dirtree_recurse</b>() manually
1320from elsewhere in the program. This gives ls -lR manual control
1321of traversal order, which is neither depth first nor breadth first but
1322instead a sort of FIFO order requried by the ls standard.</p>
1323
1324<a name="toys">
1325<h1><a href="#toys">Directory toys/</a></h1>
1326
1327<p>This directory contains command implementations. Each command is a single
1328self-contained file. Adding a new command involves adding a single
1329file, and removing a command involves removing that file. Commands use
1330shared infrastructure from the lib/ and generated/ directories.</p>
1331
1332<p>Currently there are five subdirectories under "toys/" containing "posix"
1333commands described in POSIX-2008, "lsb" commands described in the Linux
1334Standard Base 4.1, "other" commands not described by either standard,
1335"pending" commands awaiting cleanup (which default to "n" in menuconfig
1336because they don't necessarily work right yet), and "example" code showing
1337how toybox infrastructure works and providing template/skeleton files to
1338start new commands.</p>
1339
1340<p>The only difference directory location makes is which menu the command
1341shows up in during "make menuconfig", the directories are otherwise identical.
1342Note that the commands exist within a single namespace at runtime, so you can't
1343have the same command in multiple subdirectories. (The build tries to fail
1344informatively when you do that.)</p>
1345
1346<p>There is one more sub-menus in "make menuconfig" containing global
1347configuration options for toybox. This menu is defined in the top level
1348Config.in.</p>
1349
1350<p>See <a href="#adding">adding a new command</a> for details on the
1351layout of a command file.</p>
1352
1353<a name="scripts">
1354<h2>Directory scripts/</h2>
1355
1356<p>Build infrastructure. The makefile calls scripts/make.sh for "make"
1357and scripts/install.sh for "make install".</p>
1358
1359<p>There's also a test suite, "make test" calls make/test.sh, which runs all
1360the tests in make/test/*. You can run individual tests via
1361"scripts/test.sh command", or "TEST_HOST=1 scripts/test.sh command" to run
1362that test against the host implementation instead of the toybox one.</p>
1363
1364<h3>scripts/cfg2files.sh</h3>
1365
1366<p>Run .config through this filter to get a list of enabled commands, which
1367is turned into a list of files in toys via a sed invocation in the top level
1368Makefile.
1369</p>
1370
1371<h2>Directory kconfig/</h2>
1372
1373<p>Menuconfig infrastructure copied from the Linux kernel.  See the
1374Linux kernel's Documentation/kbuild/kconfig-language.txt</p>
1375
1376<!-- todo
1377
1378Better OLDTOY and multiple command explanation. From Config.in:
1379
1380<p>A command with multiple names (or multiple similar commands implemented in
1381the same .c file) should have config symbols prefixed with the name of their
1382C file. I.E. config symbol prefixes are NEWTOY() names. If OLDTOY() names
1383have config symbols they must be options (symbols with an underscore and
1384suffix) to the NEWTOY() name. (See generated/toylist.h)</p>
1385-->
1386
1387<!--#include file="footer.html" -->
1388