toybox/www/code.html

<html><head><title>toybox source code walkthrough</title></head>
<!--#include file="header.html" -->

<p><h1><a name="style" /><a href="#style">Code style</a></h1></p>

<p>The primary goal of toybox is _simple_ code. Keeping the code small is
second, with speed and lots of features coming in somewhere after that.
(For more on that, see the <a href=design.html>design</a> page.)</p>

<p>A simple implementation usually takes up fewer lines of source code,
meaning more code can fit on the screen at once, meaning the programmer can
see more of it on the screen and thus keep more if in their head at once.
This helps code auditing and thus reduces bugs. That said, sometimes being
more explicit is preferable to being clever enough to outsmart yourself:
don't be so terse your code is unreadable.</p>

<p>Toybox has an actual coding style guide over on
<a href=design.html#codestyle>the design page</a>, but in general we just
want the code to be consistent.</p>

<p><h1><a name="building" /><a href="#building">Building Toybox</a></h1></p>

<p>Toybox is configured using the
<a href=https://github.com/torvalds/linux/blob/v2.6.16/Documentation/kbuild/kconfig-language.txt>Kconfig language</a> pioneered by the Linux
kernel, and adopted by many other projects (buildroot, OpenEmbedded, etc).
This generates a ".config" file containing the selected options, which
controls which features are included when compiling toybox.</p>

<p>Each configuration option has a default value. The defaults indicate the
"maximum sane configuration", I.E. if the feature defaults to "n" then it
either isn't complete or is a special-purpose option (such as debugging
code) that isn't intended for general purpose use.</p>

<p>For a more compact human-editable version .config files, you can use the
<a href=http://landley.net/aboriginal/FAQ.html#dev_miniconfig>miniconfig</a>
format.</p>

<p>The standard build invocation is:</p>

<ul>
<li>make defconfig #(or menuconfig)</li>
<li>make</li>
<li>make install</li>
</ul>

<p>Type "make help" to see all available build options.</p>

<p>The file "configure" contains a number of environment variable definitions
which influence the build, such as specifying which compiler to use or where
to install the resulting binaries. This file is included by the build, but
accepts existing definitions of the environment variables, so it may be sourced
or modified by the developer before building and the definitions exported
to the environment will take precedence.</p>

<p>(To clarify: ".config" lists the features selected by defconfig/menuconfig,
I.E. "what to build", and "configure" describes the build and installation
environment, I.E. "how to build it".)</p>

<p>By default "make install" puts files in /usr/toybox. Adding this to the
$PATH is up to you. The environment variable $PREFIX can change the
install location, ala "PREFIX=/usr/local/bin make install".</p>

<p>If you need an unstripped (debug) version of any of these binaries,
look in generated/unstripped.</p>

<p><h1><a name="running"><a href="#running">Running a command</a></h1></p>

<h2>main</h2>

<p>The toybox main() function is at the end of main.c at the top level. It has
two possible codepaths, only one of which is configured into any given build
of toybox.</p>

<p>If CONFIG_SINGLE is selected, toybox is configured to contain only a single
command, so most of the normal setup can be skipped. In this case the
multiplexer isn't used, instead main() calls toy_singleinit() (also in main.c)
to set up global state and parse command line arguments, calls the command's
main 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
it flushes stdout (detecting error) and returns toys.exitval.</p>

<p>When CONFIG_SINGLE is not selected, main() uses basename() to find the
name it was run as, shifts its argument list one to the right so it lines up
with where the multiplexer function expects it, and calls toybox_main(). This
leverages the multiplexer command's infrastructure to find and run the
appropriate command. (A command name starting with "toybox" will
recursively call toybox_main(); you can go "./toybox toybox toybox toybox ls"
if you want to...)</p>

<h2>toybox_main</h2>

<p>The toybox_main() function is also in main,c. It handles a possible
--help option ("toybox --help ls"), prints the list of available commands if no
arguments were provided to the multiplexer (or with full path names if any
other option is provided before a command name, ala "toybox --list").
Otherwise it calls toy_exec() on its argument list.</p>

<p>Note that the multiplexer is the first entry in toy_list (the rest of the
list is sorted alphabetically to allow binary search), so toybox_main can
cheat and just grab the first entry to quickly set up its context without
searching. Since all command names go through the multiplexer at least once
in the non-TOYBOX_SINGLE case, this avoids a redundant search of
the list.</p>

<p>The toy_exec() function is also in main.c. It performs toy_find() to
perform a binary search on the toy_list array to look up the command's
entry by name and saves it in the global variable which, calls toy_init()
to parse command line arguments and set up global state (using which->options),
and calls the appropriate command's main() function (which->toy_main). On
return it flushes all pending ansi FILE * I/O, detects if stdout had an
error, and then calls xexit() (which uses toys.exitval).</p>

<p><h1><a name="infrastructure" /><a href="#infrastructure">Infrastructure</a></h1></p>

<p>The toybox source code is in following directories:</p>
<ul>
<li>The <a href="#top">top level directory</a> contains the file main.c (were
execution starts), the header file toys.h (included by every command), and
other global infrastructure.</li>
<li>The <a href="#lib">lib directory</a> contains common functions shared by
multiple commands:</li>
<ul>
<li><a href="#lib_lib">lib/lib.c</a></li>
<li><a href="#lib_xwrap">lib/xwrap.c</a></li>
<li><a href="#lib_llist">lib/llist.c</a></li>
<li><a href="#lib_args">lib/args.c</a></li>
<li><a href="#lib_dirtree">lib/dirtree.c</a></li>
</ul>
<li>The <a href="#toys">toys directory</a> contains the C files implementating
each command. Currently it contains five subdirectories categorizing the
commands: posix, lsb, other, example, and pending.</li>
<li>The <a href="#scripts">scripts directory</a> contains the build and
test infrastructure.</li>
<li>The <a href="#kconfig">kconfig directory</a> contains the configuration
infrastructure implementing menuconfig (copied from the Linux kernel).</li>
<li>The <a href="#generated">generated directory</a> contains intermediate
files generated from other parts of the source code.</li>
<li>The <a href="#tests">tests directory</a> contains the test suite.
NOSPACE=1 to allow tests to pass with diff -b</li>
</ul>

<a name="adding" />
<p><h1><a href="#adding">Adding a new command</a></h1></p>
<p>To add a new command to toybox, add a C file implementing that command to
one of the subdirectories under the toys directory.  No other files need to
be modified; the build extracts all the information it needs (such as command
line arguments) from specially formatted comments and macros in the C file.
(See the description of the <a href="#generated">"generated" directory</a>
for details.)</p>

<p>Currently there are five subdirectories under "toys", one for commands
defined by the POSIX standard, one for commands defined by the Linux Standard
Base, an "other" directory for commands not covered by an obvious standard,
a directory of example commands (templates to use when starting new commands),
and a "pending" directory of commands that need further review/cleanup
before moving to one of the other directories (run these at your own risk,
cleanup patches welcome).
These directories are just for developer convenience sorting the commands,
the directories are otherwise functionally identical. To add a new category,
create the appropriate directory with a README file in it whose first line
is the description menuconfig should use for the directory.)</p>

<p>An easy way to start a new command is copy the file "toys/example/hello.c"
to the name of the new command, and modify this copy to implement the new
command (more or less by turning every instance of "hello" into the
name of your command, updating the command line arguments, globals, and
help data, and then filling out its "main" function with code that does
something interesting).</p>

<p>You could also start with "toys/example/skeleton.c", which provides a lot
more example code (showing several variants of command line option
parsing, how to implement multiple commands in the same file, and so on).
But usually it's just more stuff to delete.</p>

<p>Here's a checklist of steps to turn hello.c into another command:</p>

<ul>
<li><p>First "cp toys/example/hello.c toys/other/yourcommand.c" and open
the new file in your preferred text editor.</p>
<ul><li><p>Note that the
name of the new file is significant: it's the name of the new command you're
adding to toybox. The build includes all *.c files under toys/*/ whose
names are a case insensitive match for an enabled config symbol. So
toys/posix/cat.c only gets included if you have "CAT=y" in ".config".</p></li>
</ul></p></li>

<li><p>Change the one line comment at the top of the file (currently
"hello.c - A hello world program") to describe your new file.</p></li>

<li><p>Change the copyright notice to your name, email, and the current
year.</p></li>

<li><p>Give a URL to the relevant standards document, where applicable.
(Sample links to SUSv4, LSB, IETF RFC, and man7.org are provided, feel free to
link to other documentation or standards as appropriate.)</p></li>

<li><p>Update the USE_YOURCOMMAND(NEWTOY(yourcommand,"blah",0)) line.
The NEWTOY macro fills out this command's <a href="#toy_list">toy_list</a>
structure.  The arguments to the NEWTOY macro are:</p>

<ol>
<li><p>the name used to run your command</p></li>
<li><p>the command line argument <a href="#lib_args">option parsing string</a> (0 if none)</p></li>
<li><p>a bitfield of TOYFLAG values
(defined in toys.h) providing additional information such as where your
command should be installed on a running system, whether to blank umask
before running, whether or not the command must run as root (and thus should
retain root access if installed SUID), and so on.</p></li>
</ol>
</li>

<li><p>Change the kconfig data (from "config YOURCOMMAND" to the end of the
comment block) to supply your command's configuration and help
information.  The uppper case config symbols are used by menuconfig, and are
also what the CFG_ and USE_() macros are generated from (see [TODO]).  The
help information here is used by menuconfig, and also by the "help" command to
describe your new command.  (See [TODO] for details.)  By convention,
unfinished commands default to "n" and finished commands default to "y",
so "make defconfig" selects all finished commands.  (Note, "finished" means
"ready to be used", not that it'll never change again.)<p>

<p>Each help block should start with a "usage: yourcommand" line explaining
any command line arguments added by this config option.  The "help" command
outputs this text, and scripts/config2help.c in the build infrastructure
collates these usage lines for commands with multiple configuration
options when producing generated/help.h.</p>
</li>

<li><p>Change the "#define FOR_hello" line to "#define FOR_yourcommand" right
before the "#include <toys.h>". (This selects the appropriate FLAG_ macros and
does a "#define TT this.yourcommand" so you can access the global variables
out of the space-saving union of structures. If you aren't using any command
flag bits and aren't defining a GLOBAL block, you can delete this line.)</p></li>

<li><p>Update the GLOBALS() macro to contain your command's global
variables. If your command has no global variables, delete this macro.</p>

<p>Variables in the GLOBALS() block are are stored in a space saving
<a href="#toy_union">union of structures</a> format, which may be accessed
using the TT macro as if TT were a global structure (so TT.membername).
If you specified two-character command line arguments in
NEWTOY(), the first few global variables will be initialized by the automatic
argument parsing logic, and the type and order of these variables must
correspond to the arguments specified in NEWTOY().
(See <a href="#lib_args">lib/args.c</a> for details.)</p>

<blockquote><p>NOTE: the GLOBALS() block creates a "this.filename" entry
in generated/globals.h. If your toys/*/filename.c does not match the first
command name, you'll need to "#define TT this.filename" yourself before
#including toys.h if you want to use TT globals</p></blockquote>
</li>

<li><p>Rename hello_main() to yourcommand_main().  This is the main() function
where execution of your command starts. Your command line options are
already sorted into this.optflags, this.optargs, this.optc, and the GLOBALS()
as appropriate by the time this function is called. (See
<a href="#lib_args">get_optflags()</a> for details.)</p></li>

<li><p>Switch on TOYBOX_DEBUG in menuconfig (toybox global settings menu)
the first time you build and run your new command. If anything is wrong
with your option string, that will give you error messages.</p>

<p>Otherwise it'll just segfault without
explanation when it falls off the end because it didn't find a matching
end parantheses for a longopt, or you put a nonexistent option in a square
bracket grouping... Since these kind of errors can only be caused by a
developer, not by end users, we don't normally want runtime checks for
them. Once you're happy with your option string, you can switch TOYBOX_DEBUG
back off.</p></li>
</ul>

<a name="headers" /><h2><a href="#headers">Headers.</a></h2>

<p>Commands are implemented as self-contained .c files, and generally don't
have their own .h files. If it's common code put it in lib/, and if it's
something like a local structure definition just put it in the command's .c
file. If it would only ever be #included from one place, inline it.
(The line between implementing multiple commands in a C file via OLDTOY()
to share infrastructure and moving that shared infrastructure to lib/ is a
judgement call. Try to figure out which is simplest.)</p>

<p>The top level toys.h should #include all the standard (posix) headers
that any command uses. (Partly this is friendly to ccache and partly this
makes the command implementations shorter.) Individual commands should only
need to include nonstandard headers that might prevent that command from
building in some context we'd care about (and thus requiring that command to
be disabled to avoid a build break).</p>

<p>Target-specific stuff (differences between compiler versions, libc versions,
or operating systems) should be confined to lib/portability.h and
lib/portability.c. (There's even some minimal compile-time environment probing
that writes data to generated/portability.h, see scripts/genconfig.sh.)</p>

<p>Only include &lt;linux/*.h&gt; headers from individual commands (not from other
headers), and only if you really need to. Data that varies per architecture
is a good reason to include a header. If you just need a couple constants
that haven't changed since the 1990's, it's ok to #define them yourself or
just use the constant inline with a comment explaining what it is. (A
#define that's only used once isn't really helping.)</p>

<p><a name="top" /><h1><a href="#top">Top level directory.</a></h1></p>

<p>This directory contains global infrastructure.</p>

<h3>toys.h</h3>
<p>Each command #includes "toys.h" as part of its standard prolog. It
may "#define FOR_commandname" before doing so to get some extra entries
specific to this command.</p>

<p>This file sucks in most of the commonly used standard #includes, so
individual files can just #include "toys.h" and not have to worry about
stdargs.h and so on.  Individual commands still need to #include
special-purpose headers that may not be present on all systems (and thus would
prevent toybox from building that command on such a system with that command
enabled).  Examples include regex support, any "linux/" or "asm/" headers, mtab
support (mntent.h and sys/mount.h), and so on.</p>

<p>The toys.h header also defines structures for most of the global variables
provided to each command by toybox_main().  These are described in
detail in the description for main.c, where they are initialized.</p>

<p>The global variables are grouped into structures (and a union) for space
savings, to more easily track the amount of memory consumed by them,
so that they may be automatically cleared/initialized as needed, and so
that access to global variables is more easily distinguished from access to
local variables.</p>

<h3>main.c</h3>
<p>Contains the main() function where execution starts, plus
common infrastructure to initialize global variables and select which command
to run.  The "toybox" multiplexer command also lives here.  (This is the
only command defined outside of the toys directory.)</p>

<p>Execution starts in main() which trims any path off of the first command
name and calls toybox_main(), which calls toy_exec(), which calls toy_find()
and toy_init() before calling the appropriate command's function from
toy_list[] (via toys.which->toy_main()).
If the command is "toybox", execution recurses into toybox_main(), otherwise
the call goes to the appropriate commandname_main() from a C file in the toys
directory.</p>

<p>The following global variables are defined in main.c:</p>
<ul>
<a name="toy_list" />
<li><p><b>struct toy_list toy_list[]</b> - array describing all the
commands currently configured into toybox.  The first entry (toy_list[0]) is
for the "toybox" multiplexer command, which runs all the other built-in commands
without symlinks by using its first argument as the name of the command to
run and the rest as that command's argument list (ala "./toybox echo hello").
The remaining entries are the commands in alphabetical order (for efficient
binary search).</p>

<p>This is a read-only array initialized at compile time by
defining macros and #including generated/newtoys.h.</p>

<p>Members of struct toy_list (defined in "toys.h") include:</p>
<ul>
<li><p>char *<b>name</b> - the name of this command.</p></li>
<li><p>void (*<b>toy_main</b>)(void) - function pointer to run this
command.</p></li>
<li><p>char *<b>options</b> - command line option string (used by
get_optflags() in lib/args.c to intialize toys.optflags, toys.optargs, and
entries in the toy's GLOBALS struct).  When this is NULL, no option
parsing is done before calling toy_main().</p></li>
<li><p>int <b>flags</b> - Behavior flags for this command.  The following flags are currently understood:</p>

<ul>
<li><b>TOYFLAG_USR</b> - Install this command under /usr</li>
<li><b>TOYFLAG_BIN</b> - Install this command under /bin</li>
<li><b>TOYFLAG_SBIN</b> - Install this command under /sbin</li>
<li><b>TOYFLAG_NOFORK</b> - This command can be used as a shell builtin.</li>
<li><b>TOYFLAG_UMASK</b> - Call umask(0) before running this command.</li>
<li><b>TOYFLAG_STAYROOT</b> - Don't drop permissions for this command if toybox is installed SUID root.</li>
<li><b>TOYFLAG_NEEDROOT</b> - This command cannot function unless run with root access.</li>
</ul>
<br>

<p>These flags are combined with | (or).  For example, to install a command
in /usr/bin, or together TOYFLAG_USR|TOYFLAG_BIN.</p>
</ul>
</li>

<li><p><b>struct toy_context toys</b> - global structure containing information
common to all commands, initializd by toy_init() and defined in "toys.h".
Members of this structure include:</p>
<ul>
<li><p>struct toy_list *<b>which</b> - a pointer to this command's toy_list
structure.  Mostly used to grab the name of the running command
(toys->which.name).</p>
</li>
<li><p>int <b>exitval</b> - Exit value of this command.  Defaults to zero.  The
error_exit() functions will return 1 if this is zero, otherwise they'll
return this value.</p></li>
<li><p>char **<b>argv</b> - "raw" command line options, I.E. the original
unmodified string array passed in to main().  Note that modifying this changes
"ps" output, and is not recommended.  This array is null terminated; a NULL
entry indicates the end of the array.</p>
<p>Most commands don't use this field, instead the use optargs, optflags,
and the fields in the GLOBALS struct initialized by get_optflags().</p>
</li>
<li><p>unsigned <b>optflags</b> - Command line option flags, set by
<a href="#lib_args">get_optflags()</a>.  Indicates which of the command line options listed in
toys->which.options occurred this time.</p>

<p>The rightmost command line argument listed in toys->which.options sets bit
1, the next one sets bit 2, and so on.  This means the bits are set in the same
order the binary digits would be listed if typed out as a string.  For example,
the option string "abcd" would parse the command line "-c" to set optflags to 2,
"-a" would set optflags to 8, and "-bd" would set optflags to 6 (4|2).</p>

<p>Only letters are relevant to optflags.  In the string "a*b:c#d", d=1, c=2,
b=4, a=8.  Punctuation after a letter initializes global variables at the
start of the GLOBALS() block (see <a href="#toy_union">union toy_union this</a>
for details).</p>

<p>The build infrastructure creates FLAG_ macros for each option letter,
corresponding to the bit position, so you can check (toys.optflags & FLAG_x)
to see if a flag was specified. (The correct set of FLAG_ macros is selected
by defining FOR_mycommand before #including toys.h. The macros live in
toys/globals.h which is generated by scripts/make.sh.)</p>

<p>For more information on option parsing, see <a href="#lib_args">get_optflags()</a>.</p>

</li>
<li><p>char **<b>optargs</b> - Null terminated array of arguments left over
after get_optflags() removed all the ones it understood.  Note: optarg[0] is
the first argument, not the command name.  Use toys.which->name for the command
name.</p></li>
<li><p>int <b>optc</b> - Optarg count, equivalent to argc but for
optargs[].<p></li>
</ul>

<a name="toy_union" />
<li><p><b>union toy_union this</b> - Union of structures containing each
command's global variables.</p>

<p>Global variables are useful: they reduce the overhead of passing extra
command line arguments between functions, they conveniently start prezeroed to
save initialization costs, and the command line argument parsing infrastructure
can also initialize global variables with its results.</p>

<p>But since each toybox process can only run one command at a time, allocating
space for global variables belonging to other commands you aren't currently
running would be wasteful.</p>

<p>Toybox handles this by encapsulating each command's global variables in
a structure, and declaring a union of those structures with a single global
instance (called "this").  The GLOBALS() macro contains the global
variables that should go in the current command's global structure.  Each
variable can then be accessed as "this.commandname.varname".
If you #defined FOR_commandname before including toys.h, the macro TT is
#defined to this.commandname so the variable can then be accessed as
"TT.variable".  See toys/hello.c for an example.</p>

<p>A command that needs global variables should declare a structure to
contain them all, and add that structure to this union.  A command should never
declare global variables outside of this, because such global variables would
allocate memory when running other commands that don't use those global
variables.</p>

<p>The first few fields of this structure can be intialized by <a href="#lib_args">get_optargs()</a>,
as specified by the options field off this command's toy_list entry.  See
the get_optargs() description in lib/args.c for details.</p>
</li>

<li><b>char toybuf[4096]</b> - a common scratch space buffer guaranteed
to start zeroed, so commands don't need to allocate/initialize their own.
Any command is free to use this, and it should never be directly referenced
by functions in lib/ (although commands are free to pass toybuf in to a
library function as an argument).</li>

<li><b>char libbuf[4096]</b> - like toybuf, but for use by common code in
lib/*.c. Commands should never directly reference libbuf, and library
could should nnever directly reference toybuf.</li>
</ul>

<p>The following functions are defined in main.c:</p>
<ul>
<li><p>struct toy_list *<b>toy_find</b>(char *name) - Return the toy_list
structure for this command name, or NULL if not found.</p></li>
<li><p>void <b>toy_init</b>(struct toy_list *which, char *argv[]) - fill out
the global toys structure, calling get_optargs() if necessary.</p></li>
<li><p>void <b>toy_exec</b>(char *argv[]) - Run a built-in command with
arguments.</p>
<p>Calls toy_find() on argv[0] (which must be just a command name
without path).  Returns if it can't find this command, otherwise calls
toy_init(), toys->which.toy_main(), and exit() instead of returning.</p>

<p>Use the library function xexec() to fall back to external executables
in $PATH if toy_exec() can't find a built-in command.  Note that toy_exec()
does not strip paths before searching for a command, so "./command" will
never match an internal command.</li>

<li><p>void <b>toybox_main</b>(void) - the main function for the multiplexer
command (I.E. "toybox").  Given a command name as its first argument, calls
toy_exec() on its arguments.  With no arguments, it lists available commands.
If the first argument starts with "-" it lists each command with its default
install path prepended.</p></li>

</ul>

<h3>Config.in</h3>

<p>Top level configuration file in a stylized variant of
<a href=http://kernel.org/doc/Documentation/kbuild/kconfig-language.txt>kconfig</a> format.  Includes generated/Config.in.</p>

<p>These files are directly used by "make menuconfig" to select which commands
to build into toybox (thus generating a .config file), and by
scripts/config2help.py to create generated/help.h.</p>

<a name="generated" />
<h1><a href="#generated">Temporary files:</a></h1>

<p>There is one temporary file in the top level source directory:</p>
<ul>
<li><p><b>.config</b> - Configuration file generated by kconfig, indicating
which commands (and options to commands) are currently enabled.  Used
to make generated/config.h and determine which toys/*/*.c files to build.</p>

<p>You can create a human readable "miniconfig" version of this file using
<a href=http://landley.net/aboriginal/new_platform.html#miniconfig>these
instructions</a>.</p>
</li>
</ul>

<p><h2>Directory generated/</h2></p>

<p>The remaining temporary files live in the "generated/" directory,
which is for files generated at build time from other source files.</p>

<ul>
<li><p><b>generated/Config.in</b> - Kconfig entries for each command, included
from the top level Config.in. The help text here is used to generate
help.h.</p>

<p>Each command has a configuration entry with an upper case version of
the command name. Options to commands start with the command
name followed by an underscore and the option name. Global options are attached
to the "toybox" command, and thus use the prefix "TOYBOX_".  This organization
is used by scripts/cfg2files to select which toys/*/*.c files to compile for a
given .config.</p>
</li>

<li><p><b>generated/config.h</b> - list of CFG_SYMBOL and USE_SYMBOL() macros,
generated from .config by a sed invocation in scripts/make.sh.</p>

<p>CFG_SYMBOL is a comple time constant set to 1 for enabled symbols and 0 for
disabled symbols. This allows the use of normal if() statements to remove
code at compile time via the optimizer's dead code elimination (which removes
from the binary any code that cannot be reached). This saves space without
cluttering the code with #ifdefs or leading to configuration dependent build
breaks. (See the 1992 Usenix paper
<a href=http://doc.cat-v.org/henry_spencer/ifdef_considered_harmful.pdf>#ifdef
Considered Harmful</a> for more information.)</p>

<p>When you can't entirely avoid an #ifdef, the USE_SYMBOL(code) macro
provides a less intrusive alternative, evaluating to the code in parentheses
when the symbol is enabled, and nothing when the symbol is disabled. This
is most commonly used around NEWTOY() declarations (so only the enabled
commands show up in toy_list), and in option strings. This can also be used
for things like varargs or structure members which can't always be
eliminated by a simple test on CFG_SYMBOL. Remember, unlike CFG_SYMBOL
this is really just a variant of #ifdef, and can still result in configuration
dependent build breaks. Use with caution.</p>
</li>

<li><p><b>generated/flags.h</b> - FLAG_? macros indicating which command
line options were seen. The option parsing in lib/args.c sets bits in
toys.optflags, which can be tested by anding with the appropriate FLAG_
macro. (Bare longopts, which have no corresponding short option, will
have the longopt name after FLAG_. All others use the single letter short
option.)</p>

<p>To get the appropriate macros for your command, #define FOR_commandname
before #including toys.h. To switch macro sets (because you have an OLDTOY()
with different options in the same .c file), #define CLEANUP_oldcommand
and also #define FOR_newcommand, then #include "generated/flags.h" to switch.
</p>
</li>

<li><p><b>generated/globals.h</b> -
Declares structures to hold the contents of each command's GLOBALS(),
and combines them into "global_union this". (Yes, the name was
chosen to piss off C++ developers who think that C
is merely a subset of C++, not a language in its own right.)</p>

<p>The union reuses the same memory for each command's global struct:
since only one command's globals are in use at any given time, collapsing
them together saves space. The headers #define TT to the appropriate
"this.commandname", so you can refer to the current command's global
variables out of "this" as TT.variablename.</p>

<p>The globals start zeroed, and the first few are filled out by the
lib/args.c argument parsing code called from main.c.</p>
</li>

<li><p><b>toys/help.h</b> - Help strings for use by the "help" command and
--help options. This file #defines a help_symbolname string for each
symbolname, but only the symbolnames matching command names get used
by show_help() in lib/help.c to display help for commands.</p>

<p>This file is created by scripts/make.sh, which compiles scripts/config2help.c
into the binary generated/config2help, and then runs it against the top
level .config and Config.in files to extract the help text from each config
entry and collate together dependent options.</p>

<p>This file contains help text for all commands, regardless of current
configuration, but only the ones currently enabled in the .config file
wind up in the help_data[] array, and only the enabled dependent options
have their help text added to the command they depend on.</p>
</li>

<li><p><b>generated/newtoys.h</b> -
All the NEWTOY() and OLDTOY() macros from toys/*/*.c. The "toybox" multiplexer
is the first entry, the rest are in alphabetical order. Each line should be
inside an appropriate USE_ macro, so code that #includes this file only sees
the currently enabled commands.</p>

<p>By #definining NEWTOY() to various things before #including this file,
it may be used to create function prototypes (in toys.h), initialize the
help_data array (in lib/help.c),  initialize the toy_list array (in main.c,
the alphabetical order lets toy_find() do a binary search, the exception to
the alphabetical order lets it use the multiplexer without searching), and so
on.  (It's even used to initialize the NEED_OPTIONS macro, which produces a 1
or 0 for each command using command line option parsing, which is ORed together
to allow compile-time dead code elimination to remove the whole of
lib/args.c if nothing currently enabled is using it.)<p>

<p>Each NEWTOY and OLDTOY macro contains the command name, command line
option string (telling lib/args.c how to parse command line options for
this command), recommended install location, and miscelaneous data such
as whether this command should retain root permissions if installed suid.</p>
</li>

<li><p><b>toys/oldtoys.h</b> - Macros with the command line option parsing
string for each NEWTOY. This allows an OLDTOY that's just an alias for an
existing command to refer to the existing option string instead of
having to repeat it.</p>
</li>
</ul>

<a name="lib">
<h2>Directory lib/</h2>

<p>TODO: document lots more here.</p>

<p>lib: getmountlist(), error_msg/error_exit, xmalloc(),
strlcpy(), xexec(), xopen()/xread(), xgetcwd(), xabspath(), find_in_path(),
itoa().</p>


<a name="lib_xwrap"><h3>lib/xwrap.c</h3>

<p>Functions prefixed with the letter x call perror_exit() when they hit
errors, to eliminate common error checking. This prints an error message
and the strerror() string for the errno encountered.</p>

<p>We replaced exit(), _exit(), and atexit() with xexit(), _xexit(), and
sigatexit(). This gives _xexit() the option to siglongjmp(toys.rebound, 1)
instead of exiting, lets xexit() report stdout flush failures to stderr
and change the exit code to indicate error, lets our toys.exit function
change happen for signal exit paths and lets us remove the functions
after we've called them.</p>

<p>You can intercept our exit by assigning a sigsetjmp/siglongjmp buffer to
toys.rebound (set it back to zero to restore the default behavior).
If you do this, cleaning up resource leaks is your problem.</p>

<ul>
<li><b>void xstrncpy(char *dest, char *src, size_t size)</b></li>
<li><p><b><p>void _xexit(void)</b></p>
<p>Calls siglongjmp(toys.rebound, 1), or else _exit(toys.exitval). This
lets you ignore errors with the NO_EXIT() macro wrapper, or intercept
them with WOULD_EXIT().</p>
<li><b><p>void xexit(void)</b></p>
<p>Calls toys.xexit functions (if any) and flushes stdout/stderr (reporting
failure to write to stdout both to stderr and in the exit code), then
calls _xexit().</p>
</li>
<li><b>void *xmalloc(size_t size)</b></li>
<li><b>void *xzalloc(size_t size)</b></li>
<li><b>void *xrealloc(void *ptr, size_t size)</b></li>
<li><b>char *xstrndup(char *s, size_t n)</b></li>
<li><b>char *xstrdup(char *s)</b></li>
<li><b>char *xmprintf(char *format, ...)</b></li>
<li><b>void xprintf(char *format, ...)</b></li>
<li><b>void xputs(char *s)</b></li>
<li><b>void xputc(char c)</b></li>
<li><b>void xflush(void)</b></li>
<li><b>pid_t xfork(void)</b></li>
<li><b>void xexec_optargs(int skip)</b></li>
<li><b>void xexec(char **argv)</b></li>
<li><b>pid_t xpopen(char **argv, int *pipes)</b></li>
<li><b>int xpclose(pid_t pid, int *pipes)</b></li>
<li><b>void xaccess(char *path, int flags)</b></li>
<li><b>void xunlink(char *path)</b></li>
<li><p><b>int xcreate(char *path, int flags, int mode)<br />
int xopen(char *path, int flags)</b></p>

<p>The xopen() and xcreate() functions open an existing file (exiting if
it's not there) and create a new file (exiting if it can't).</p>

<p>They default to O_CLOEXEC so the filehandles aren't passed on to child
processes. Feed in O_CLOEXEC to disable this.</p>
</li>
<li><p><b>void xclose(int fd)</b></p>

<p>Because NFS is broken, and won't necessarily perform the requested
operation (and report the error) until you close the file. Of course, this
being NFS, it's not guaranteed to report the error there either, but it
_can_.</p>

<p>Nothing else ever reports an error on close, everywhere else it's just a
VFS operation freeing some resources. NFS is _special_, in a way that
other network filesystems like smbfs and v9fs aren't..</p>
</li>
<li><b>int xdup(int fd)</b></li>
<li><p><b>size_t xread(int fd, void *buf, size_t len)</b></p>

<p>Can return 0, but not -1.</p>
</li>
<li><p><b>void xreadall(int fd, void *buf, size_t len)</b></p>

<p>Reads the entire len-sized buffer, retrying to complete short
reads. Exits if it can't get enough data.</p></li>

<li><p><b>void xwrite(int fd, void *buf, size_t len)</b></p>

<p>Retries short writes, exits if can't write the entire buffer.</p></li>

<li><b>off_t xlseek(int fd, off_t offset, int whence)</b></li>
<li><b>char *xgetcwd(void)</b></li>
<li><b>void xstat(char *path, struct stat *st)</b></li>
<li><p><b>char *xabspath(char *path, int exact) </b></p>

<p>After several years of
<a href=http://landley.net/notes-2007.html#18-06-2007>wrestling</a>
<a href=http://landley.net/notes-2008.html#19-01-2008>with</a> realpath(),
I broke down and <a href=http://landley.net/notes-2012.html#20-11-2012>wrote
my own</a> implementation that doesn't use the one in libc. As I explained:

<blockquote><p>If the path ends with a broken link,
readlink -f should show where the link points to, not where the broken link
lives. (The point of readlink -f is "if I write here, where would it attempt
to create a file".) The problem is, realpath() returns NULL for a path ending
with a broken link, and I can't beat different behavior out of code locked
away in libc.</p></blockquote>

<p>
</li>
<li><b>void xchdir(char *path)</b></li>
<li><b>void xchroot(char *path)</b></li>

<li><p><b>struct passwd *xgetpwuid(uid_t uid)<br />
struct group *xgetgrgid(gid_t gid)<br />
struct passwd *xgetpwnam(char *name)</b></p>
</li>

<li><b>void xsetuser(struct passwd *pwd)</b></li>
<li><b>char *xreadlink(char *name)</b></li>
<li><b>char *xreadfile(char *name, char *buf, off_t len)</b></li>
<li><b>int xioctl(int fd, int request, void *data)</b></li>
<li><b>void xpidfile(char *name)</b></li>
<li><b>void xsendfile(int in, int out)</b></li>
<li><b>long xparsetime(char *arg, long units, long *fraction)</b></li>
<li><b>void xregcomp(regex_t *preg, char *regex, int cflags)</b></li>
</ul>

<a name="lib_lib"><h3>lib/lib.c</h3>
<p>Eight gazillion common functions, see lib/lib.h for the moment:</p>

<h3>lib/portability.h</h3>

<p>This file is automatically included from the top of toys.h, and smooths
over differences between platforms (hardware targets, compilers, C libraries,
operating systems, etc).</p>

<p>This file provides SWAP macros (SWAP_BE16(x) and SWAP_LE32(x) and so on).</p>

<p>A macro like SWAP_LE32(x) means "The value in x is stored as a little
endian 32 bit value, so perform the translation to/from whatever the native
32-bit format is".  You do the swap once on the way in, and once on the way
out. If your target is already little endian, the macro is a NOP.</p>

<p>The SWAP macros come in BE and LE each with 16, 32, and 64 bit versions.
In each case, the name of the macro refers to the _external_ representation,
and converts to/from whatever your native representation happens to be (which
can vary depending on what you're currently compiling for).</p>

<a name="lib_llist"><h3>lib/llist.c</h3>

<p>Some generic single and doubly linked list functions, which take
advantage of a couple properties of C:</p>

<ul>
<li><p>Structure elements are laid out in memory in the order listed, and
the first element has no padding. This means you can always treat (typecast)
a pointer to a structure as a pointer to the first element of the structure,
even if you don't know anything about the data following it.</p></li>

<li><p>An array of length zero at the end of a structure adds no space
to the sizeof() the structure, but if you calculate how much extra space
you want when you malloc() the structure it will be available at the end.
Since C has no bounds checking, this means each struct can have one variable
length array.</p></li>
</ul>

<p>Toybox's list structures always have their <b>next</b> pointer as
the first entry of each struct, and singly linked lists end with a NULL pointer.
This allows generic code to traverse such lists without knowing anything
else about the specific structs composing them: if your pointer isn't NULL
typecast it to void ** and dereference once to get the next entry.</p>

<p><b>lib/lib.h</b> defines three structure types:</p>
<ul>
<li><p><b>struct string_list</b> - stores a single string (<b>char str[0]</b>),
memory for which is allocated as part of the node. (I.E. llist_traverse(list,
free); can clean up after this type of list.)</p></li>

<li><p><b>struct arg_list</b> - stores a pointer to a single string
(<b>char *arg</b>) which is stored in a separate chunk of memory.</p></li>

<li><p><b>struct double_list</b> - has a second pointer (<b>struct double_list
*prev</b> along with a <b>char *data</b> for payload.</p></li>
</ul>

<b>List Functions</b>

<ul>
<li><p>void *<b>llist_pop</b>(void **list) - advances through a list ala
<b>node = llist_pop(&list);</b>  This doesn't modify the list contents,
but does advance the pointer you feed it (which is why you pass the _address_
of that pointer, not the pointer itself).</p></li>

<li><p>void <b>llist_traverse</b>(void *list, void (*using)(void *data)) -
iterate through a list calling a function on each node.</p></li>

<li><p>struct double_list *<b>dlist_add</b>(struct double_list **llist, char *data)
- append an entry to a circular linked list.
This function allocates a new struct double_list wrapper and returns the
pointer to the new entry (which you can usually ignore since it's llist->prev,
but if llist was NULL you need it). The argument is the ->data field for the
new node.</p></li>
<ul><li><p>void <b>dlist_add_nomalloc</b>(struct double_list **llist,
struct double_list *new) - append existing struct double_list to
list, does not allocate anything.</p></li></ul>
</ul>

<b>List code trivia questions:</b>

<ul>
<li><p><b>Why do arg_list and double_list contain a char * payload instead of
a void *?</b> - Because you always have to typecast a void * to use it, and
typecasting a char * does no harm. Since strings are the most common
payload, and doing math on the pointer ala
"(type *)(ptr+sizeof(thing)+sizeof(otherthing))" requires ptr to be char *
anyway (at least according to the C standard), defaulting to char * saves
a typecast.</p>
</li>

<li><p><b>Why do the names ->str, ->arg, and ->data differ?</b> - To force
you to keep track of which one you're using, calling free(node->str) would
be bad, and _failing_ to free(node->arg) leaks memory.</p></li>

<li><p><b>Why does llist_pop() take a void * instead of void **?</b> -
because the stupid compiler complains about "type punned pointers" when
you typecast and dereference on the same line,
due to insane FSF developers hardwiring limitations of their optimizer
into gcc's warning system. Since C automatically typecasts any other
pointer type to and from void *, the current code works fine. It's sad that it
won't warn you if you forget the &, but the code crashes pretty quickly in
that case.</p></li>

<li><p><b>How do I assemble a singly-linked-list in order?</b> - use
a double_list, dlist_add() your entries, and then call dlist_terminate(list)
to break the circle when done (turning the last ->next and the first ->prev
into NULLs).</p>
</ul>

<a name="lib_args"><h3>lib/args.c</h3>

<p>Toybox's main.c automatically parses command line options before calling the
command's main function. Option parsing starts in get_optflags(), which stores
results in the global structures "toys" (optflags and optargs) and "this".</p>

<p>The option parsing infrastructure stores a bitfield in toys.optflags to
indicate which options the current command line contained, and defines FLAG
macros code can use to check whether each argument's bit is set. Arguments
attached to those options are saved into the command's global structure
("this"). Any remaining command line arguments are collected together into
the null-terminated array toys.optargs, with the length in toys.optc. (Note
that toys.optargs does not contain the current command name at position zero,
use "toys.which->name" for that.) The raw command line arguments get_optflags()
parsed are retained unmodified in toys.argv[].</p>

<p>Toybox's option parsing logic is controlled by an "optflags" string, using
a format reminiscent of getopt's optargs but with several important differences.
Toybox does not use the getopt()
function out of the C library, get_optflags() is an independent implementation
which doesn't permute the original arguments (and thus doesn't change how the
command is displayed in ps and top), and has many features not present in
libc optargs() (such as the ability to describe long options in the same string
as normal options).</p>

<p>Each command's NEWTOY() macro has an optflags string as its middle argument,
which sets toy_list.options for that command to tell get_optflags() what
command line arguments to look for, and what to do with them.
If a command has no option
definition string (I.E. the argument is NULL), option parsing is skipped
for that command, which must look at the raw data in toys.argv to parse its
own arguments. (If no currently enabled command uses option parsing,
get_optflags() is optimized out of the resulting binary by the compiler's
--gc-sections option.)</p>

<p>You don't have to free the option strings, which point into the environment
space (I.E. the string data is not copied). A TOYFLAG_NOFORK command
that uses the linked list type "*" should free the list objects but not
the data they point to, via "llist_free(TT.mylist, NULL);". (If it's not
NOFORK, exit() will free all the malloced data anyway unless you want
to implement a CONFIG_TOYBOX_FREE cleanup for it.)</p>

<h4>Optflags format string</h4>

<p>Note: the optflags option description string format is much more
concisely described by a large comment at the top of lib/args.c.</p>

<p>The general theory is that letters set optflags, and punctuation describes
other actions the option parsing logic should take.</p>

<p>For example, suppose the command line <b>command -b fruit -d walrus -a 42</b>
is parsed using the optflags string "<b>a#b:c:d</b>".  (I.E.
toys.which->options="a#b:c:d" and argv = ["command", "-b", "fruit", "-d",
"walrus", "-a", "42"]).  When get_optflags() returns, the following data is
available to command_main():

<ul>
<li><p>In <b>struct toys</b>:
<ul>
<li>toys.optflags = 13; // FLAG_a = 8 | FLAG_b = 4 | FLAG_d = 1</li>
<li>toys.optargs[0] = "walrus"; // leftover argument</li>
<li>toys.optargs[1] = NULL; // end of list</li>
<li>toys.optc = 1; // there was 1 leftover argument</li>
<li>toys.argv[] = {"-b", "fruit", "-d", "walrus", "-a", "42"}; // The original command line arguments
</ul>
<p></li>

<li><p>In <b>union this</b> (treated as <b>long this[]</b>):
<ul>
<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>
<li>this[1] = (long)"fruit"; // argument to -b</li>
<li>this[2] = 42; // argument to -a</li>
</ul>
</p></li>
</ul>

<p>If the command's globals are:</p>

<blockquote><pre>
GLOBALS(
	char *c;
	char *b;
	long a;
)
</pre></blockquote>

<p>That would mean TT.c == NULL, TT.b == "fruit", and TT.a == 42.  (Remember,
each entry that receives an argument must be a long or pointer, to line up
with the array position.  Right to left in the optflags string corresponds to
top to bottom in GLOBALS().</p>

<p>Put globals not filled out by the option parsing logic at the end of the
GLOBALS block. Common practice is to list the options one per line (to
make the ordering explicit, first to last in globals corresponds to right
to left in the option string), then leave a blank line before any non-option
globals.</p>

<p><b>long toys.optflags</b></p>

<p>Each option in the optflags string corresponds to a bit position in
toys.optflags, with the same value as a corresponding binary digit.  The
rightmost argument is (1<<0), the next to last is (1<<1) and so on.  If
the option isn't encountered while parsing argv[], its bit remains 0.</p>

<p>Each option -x has a FLAG_x macro for the command letter. Bare --longopts
with no corresponding short option have a FLAG_longopt macro for the long
optionname. Commands enable these macros by #defining FOR_commandname before
#including <toys.h>. When multiple commands are implemented in the same
source file, you can switch flag contexts later in the file by
#defining CLEANUP_oldcommand and #defining FOR_newcommand, then
#including <generated/flags.h>.</p>

<p>Options disabled in the current configuration (wrapped in
a USE_BLAH() macro for a CONFIG_BLAH that's switched off) have their
corresponding FLAG macro set to zero, so code checking them ala
if (toys.optargs & FLAG_x) gets optimized out via dead code elimination.
#defining FORCE_FLAGS when switching flag context disables this
behavior: the flag is never zero even if the config is disabled. This
allows code shared between multiple commands to use the same flag
values, as long as the common flags match up right to left in both option
strings.</p>

<p>For example,
the optflags string "abcd" would parse the command line argument "-c" to set
optflags to 2, "-a" would set optflags to 8, "-bd" would set optflags to
6 (I.E. 4|2), and "-a -c" would set optflags to 10 (2|8). To check if -c
was encountered, code could test: if (toys.optflags & FLAG_c) printf("yup");
(See the toys/examples directory for more.)</p>

<p>Only letters are relevant to optflags, punctuation is skipped: in the
string "a*b:c#d", d=1, c=2, b=4, a=8. The punctuation after a letter
usually indicate that the option takes an argument.</p>

<p>Since toys.optflags is an unsigned int, it only stores 32 bits. (Which is
the amount a long would have on 32-bit platforms anyway; 64 bit code on
32 bit platforms is too expensive to require in common code used by almost
all commands.) Bit positions beyond the 1<<31 aren't recorded, but
parsing higher options can still set global variables.</p>

<p><b>Automatically setting global variables from arguments (union this)</b></p>

<p>The following punctuation characters may be appended to an optflags
argument letter, indicating the option takes an additional argument:</p>

<ul>
<li><b>:</b> - plus a string argument, keep most recent if more than one.</li>
<li><b>*</b> - plus a string argument, appended to a linked list.</li>
<li><b>@</b> - plus an occurrence counter (stored in a long)</li>
<li><b>#</b> - plus a signed long argument.
<li><b>-</b> - plus a signed long argument defaulting to negative (start argument with + to force a positive value).</li>
<li><b>.</b> - plus a floating point argument (if CFG_TOYBOX_FLOAT).</li>
<ul>The following can be appended to a float or double:
<li><b>&lt;123</b> - error if argument is less than this</li>
<li><b>&gt;123</b> - error if argument is greater than this</li>
<li><b>=123</b> - default value if argument not supplied</li>
</ul>
</ul>

<p><b>GLOBALS</b></p>

<p>Options which have an argument fill in the corresponding slot in the global
union "this" (see generated/globals.h), treating it as an array of longs
with the rightmost saved in this[0].  As described above, using "a*b:c#d",
"-c 42" would set this[0] = 42; and "-b 42" would set this[1] = "42"; each
slot is left NULL if the corresponding argument is not encountered.</p>

<p>This behavior is useful because the LP64 standard ensures long and pointer
are the same size. C99 guarantees structure members will occur in memory
in the same order they're declared, and that padding won't be inserted between
consecutive variables of register size.  Thus the first few entries can
be longs or pointers corresponding to the saved arguments.</p>

<p>The main downside is that numeric arguments ("#" and "-" format)
are limited to +- 2 billion on 32 bit platforms (the "truncate -s 8G"
problem), because long is only 64 bits on 64 bit hosts, so the capabilities
of some tools differ when built in 32 bit vs 64 bit mode. Fixing this
kind of ugly and even embedded designs are slowly moving to 64 bits,
so our current plan is to document the problem and wait it out. (If
"x32 mode" and similar becomes popular enough, we may revisit this
decision.)</p>

<p>See toys/example/*.c for longer examples of parsing options into the
GLOBALS block.</p>

<p><b>char *toys.optargs[]</b></p>

<p>Command line arguments in argv[] which are not consumed by option parsing
(I.E. not recognized either as -flags or arguments to -flags) will be copied
to toys.optargs[], with the length of that array in toys.optc.
(When toys.optc is 0, no unrecognized command line arguments remain.)
The order of entries is preserved, and as with argv[] this new array is also
terminated by a NULL entry.</p>

<p>Option parsing can require a minimum or maximum number of optargs left
over, by adding "<1" (read "at least one") or ">9" ("at most nine") to the
start of the optflags string.</p>

<p>The special argument "--" terminates option parsing, storing all remaining
arguments in optargs.  The "--" itself is consumed.</p>

<p><b>Other optflags control characters</b></p>

<p>The following characters may occur at the start of each command's
optflags string, before any options that would set a bit in toys.optflags:</p>

<ul>
<li><b>^</b> - stop at first nonoption argument (for nice, xargs...)</li>
<li><b>?</b> - allow unknown arguments (pass non-option arguments starting
with - through to optargs instead of erroring out).</li>
<li><b>&amp;</b> - the first argument has imaginary dash (ala tar/ps.  If given twice, all arguments have imaginary dash.)</li>
<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>
<li><b>&gt;</b> - must be followed by a decimal digit indicating at most this many leftover arguments allowed (default MAX_INT)</li>
</ul>

<p>The following characters may be appended to an option character, but do
not by themselves indicate an extra argument should be saved in this[].
(Technically any character not recognized as a control character sets an
optflag, but letters are never control characters.)</p>

<ul>
<li><b>^</b> - stop parsing options after encountering this option, everything else goes into optargs.</li>
<li><b>|</b> - this option is required.  If more than one marked, only one is required.</li>
</ul>

<p>The following may be appended to a float or double:</p>

<ul>
<li><b>&lt;123</b> - error if argument is less than this</li>
<li><b>&gt;123</b> - error if argument is greater than this</li>
<li><b>=123</b> - default value if argument not supplied</li>
</ul>

<p>Option parsing only understands <>= after . when CFG_TOYBOX_FLOAT
is enabled. (Otherwise the code to determine where floating point constants
end drops out.  When disabled, it can reserve a global data slot for the
argument so offsets won't change, but will never fill it out.) You can handle
this by using the USE_BLAH() macros with C string concatenation, ala:</p>

<blockquote>"abc." USE_TOYBOX_FLOAT("<1.23>4.56=7.89") "def"</blockquote>

<p><b>--longopts</b></p>

<p>The optflags string can contain long options, which are enclosed in
parentheses. They may be appended to an existing option character, in
which case the --longopt is a synonym for that option, ala "a:(--fred)"
which understands "-a blah" or "--fred blah" as synonyms.</p>

<p>Longopts may also appear before any other options in the optflags string,
in which case they have no corresponding short argument, but instead set
their own bit based on position. So for "(walrus)#(blah)xy:z", "command
--walrus 42" would set toys.optflags = 16 (-z = 1, -y = 2, -x = 4, --blah = 8)
and would assign this[1] = 42;</p>

<p>A short option may have multiple longopt synonyms, "a(one)(two)", but
each "bare longopt" (ala "(one)(two)abc" before any option characters)
always sets its own bit (although you can group them with +X).</p>

<p>Only bare longopts have a FLAG_ macro with the longopt name
(ala --fred would #define FLAG_fred). Other longopts use the short
option's FLAG macro to test the toys.optflags bit.</p>

<p>Options with a semicolon ";" after their data type can only set their
corresponding GLOBALS() entry via "--longopt=value". For example, option
string "x(boing): y" would set TT.x if it saw "--boing=value", but would
treat "--boing value" as setting FLAG_x in toys.optargs, leaving TT.x NULL,
and keeping "value" in toys.optargs[]. (This lets "ls --color" and
"ls --color=auto" both work.)</p>

<p><b>[groups]</b></p>

<p>At the end of the option string, square bracket groups can define
relationships between existing options. (This only applies to short
options, bare --longopts can't participate.)</p>

<p>The first character of the group defines the type, the remaining
characters are options it applies to:</p>

<ul>
<li><b>-</b> - Exclusive, switch off all others in this group.</li>
<li><b>+</b> - Inclusive, switch on all others in this group.</li>
<li><b>!</b> - Error, fail if more than one defined.</li>
</ul>

<p>So "abc[-abc]" means -ab = -b, -ba = -a, -abc = -c. "abc[+abc]"
means -ab=-abc, -c=-abc, and "abc[!abc] means -ab calls error_exit("no -b
with -a"). Note that [-] groups clear the GLOBALS option slot of
options they're switching back off, but [+] won't set options it didn't see
(just the optflags).</p>

<p><b>whitespace</b></p>

<p>Arguments may occur with or without a space (I.E. "-a 42" or "-a42").
The command line argument "-abc" may be interepreted many different ways:
the optflags string "cba" sets toys.optflags = 7, "c:ba" sets toys.optflags=4
and saves "ba" as the argument to -c, and "cb:a" sets optflags to 6 and saves
"c" as the argument to -b.</p>

<p>Note that &amp; changes whitespace handling, so that the command line
"tar cvfCj outfile.tar.bz2 topdir filename" is parsed the same as
"tar filename -c -v -j -f outfile.tar.bz2 -C topdir". Note that "tar -cvfCj
one two three" would equal "tar -c -v -f Cj one two three". (This matches
historical usage.)</p>

<p>Appending a space to the option in the option string ("a: b") makes it
require a space, I.E. "-ab" is interpreted as "-a" "-b". That way "kill -stop"
differs from "kill -s top".</p>

<p>Appending ; to a longopt in the option string makes its argument optional,
and only settable with =, so in ls "(color):;" can accept "ls --color" and
"ls --color=auto" without complaining that the first has no argument.</p>

<a name="lib_dirtree"><h3>lib/dirtree.c</h3>

<p>The directory tree traversal code should be sufficiently generic
that commands never need to use readdir(), scandir(), or the fts.h family
of functions.</p>

<p>These functions do not call chdir() or rely on PATH_MAX. Instead they
use openat() and friends, using one filehandle per directory level to
recurse into subdirectories. (I.E. they can descend 1000 directories deep
if setrlimit(RLIMIT_NOFILE) allows enough open filehandles, and the default
in /proc/self/limits is generally 1024.)</p>

<p>There are two main ways to use dirtree: 1) assemble a tree of nodes
representing a snapshot of directory state and traverse them using the
->next and ->child pointers, or 2) traverse the tree calling a callback
function on each entry, and freeing its node afterwards. (You can also
combine the two, using the callback as a filter to determine which nodes
to keep.)</p>

<p>The basic dirtree functions are:</p>

<ul>
<li><p><b>struct dirtree *dirtree_read(char *path, int (*callback)(struct
dirtree node))</b> - recursively read files and directories, calling
callback() on each, and returning a tree of saved nodes (if any).
If path doesn't exist, returns DIRTREE_ABORTVAL. If callback is NULL,
returns a single node at that path.</p>

<li><p><b>dirtree_notdotdot(struct dirtree *new)</b> - standard callback
which discards "." and ".." entries and returns DIRTREE_SAVE|DIRTREE_RECURSE
for everything else. Used directly, this assembles a snapshot tree of
the contents of this directory and its subdirectories
to be processed after dirtree_read() returns (by traversing the
struct dirtree's ->next and ->child pointers from the returned root node).</p>

<li><p><b>dirtree_path(struct dirtree *node, int *plen)</b> - malloc() a
string containing the path from the root of this tree to this node. If
plen isn't NULL then *plen is how many extra bytes to malloc at the end
of string.</p></li>

<li><p><b>dirtree_parentfd(struct dirtree *node)</b> - return fd of
directory containing this node, for use with openat() and such.</p></li>
</ul>

<p>The <b>dirtree_read()</b> function is the standard way to start
directory traversal. It takes two arguments: a starting path for
the root of the tree, and a callback function. The callback() is called
on each directory entry, its argument is a fully populated
<b>struct dirtree *</b> (from lib/lib.h) describing the node, and its
return value tells the dirtree infrastructure what to do next.</p>

<p>(There's also a three argument version,
<b>dirtree_flagread(char *path, int flags, int (*callback)(struct
dirtree node))</b>, which lets you apply flags like DIRTREE_SYMFOLLOW and
DIRTREE_SHUTUP to reading the top node, but this only affects the top node.
Child nodes use the flags returned by callback().</p>

<p><b>struct dirtree</b></p>

<p>Each struct dirtree node contains <b>char name[]</b> and <b>struct stat
st</b> entries describing a file, plus a <b>char *symlink</b>
which is NULL for non-symlinks.</p>

<p>During a callback function, the <b>int dirfd</b> field of directory nodes
contains a directory file descriptor (for use with the openat() family of
functions). This isn't usually used directly, intstead call dirtree_parentfd()
on the callback's node argument. The <b>char again</b> field is 0 for the
first callback on a node, and 1 on the second callback (triggered by returning
DIRTREE_COMEAGAIN on a directory, made after all children have been processed).
</p>

<p>Users of this code may put anything they like into the <b>long extra</b>
field. For example, "cp" and "mv" use this to store a dirfd for the destination
directory (and use DIRTREE_COMEAGAIN to get the second callback so they can
close(node->extra) to avoid running out of filehandles).
This field is not directly used by the dirtree code, and
thanks to LP64 it's large enough to store a typecast pointer to an
arbitrary struct.</p>

<p>The return value of the callback combines flags (with boolean or) to tell
the traversal infrastructure how to behave:</p>

<ul>
<li><p><b>DIRTREE_SAVE</b> - Save this node, assembling a tree. (Without
this the struct dirtree is freed after the callback returns. Filtering out
siblings is fine, but discarding a parent while keeping its child leaks
memory.)</p></li>
<li><p><b>DIRTREE_ABORT</b> - Do not examine any more entries in this
directory. (Does not propagate up tree: to abort entire traversal,
return DIRTREE_ABORT from parent callbacks too.)</p></li>
<li><p><b>DIRTREE_RECURSE</b> - Examine directory contents. Ignored for
non-directory entries. The remaining flags only take effect when
recursing into the children of a directory.</p></li>
<li><p><b>DIRTREE_COMEAGAIN</b> - Call the callback on this node a second time
after examining all directory contents, allowing depth-first traversal.
On the second call, dirtree->again is nonzero.</p></li>
<li><p><b>DIRTREE_SYMFOLLOW</b> - follow symlinks when populating children's
<b>struct stat st</b> (by feeding a nonzero value to the symfollow argument of
dirtree_add_node()), which means DIRTREE_RECURSE treats symlinks to
directories as directories. (Avoiding infinite recursion is the callback's
problem: the non-NULL dirtree->symlink can still distinguish between
them. The "find" command follows ->parent up the tree to the root node
each time, checking to make sure that stat's dev and inode pair don't
match any ancestors.)</p></li>
</ul>

<p>Each struct dirtree contains three pointers (next, parent, and child)
to other struct dirtree.</p>

<p>The <b>parent</b> pointer indicates the directory
containing this entry; even when not assembling a persistent tree of
nodes the parent entries remain live up to the root of the tree while
child nodes are active. At the top of the tree the parent pointer is
NULL, meaning the node's name[] is either an absolute path or relative
to cwd. The function dirtree_parentfd() gets the directory file descriptor
for use with openat() and friends, returning AT_FDCWD at the top of tree.</p>

<p>The <b>child</b> pointer points to the first node of the list of contents of
this directory. If the directory contains no files, or the entry isn't
a directory, child is NULL.</p>

<p>The <b>next</b> pointer indicates sibling nodes in the same directory as this
node, and since it's the first entry in the struct the llist.c traversal
mechanisms work to iterate over sibling nodes. Each dirtree node is a
single malloc() (even char *symlink points to memory at the end of the node),
so llist_free() works but its callback must descend into child nodes (freeing
a tree, not just a linked list), plus whatever the user stored in extra.</p>

<p>The <b>dirtree_flagread</b>() function is a simple wrapper, calling <b>dirtree_add_node</b>()
to create a root node relative to the current directory, then calling
<b>dirtree_handle_callback</b>() on that node (which recurses as instructed by the callback
return flags). The flags argument primarily lets you
control whether or not to follow symlinks to the root node; symlinks
listed on the command line are often treated differently than symlinks
encountered during recursive directory traversal.

<p>The ls command not only bypasses this wrapper, but never returns
<b>DIRTREE_RECURSE</b> from the callback, instead calling <b>dirtree_recurse</b>() manually
from elsewhere in the program. This gives ls -lR manual control
of traversal order, which is neither depth first nor breadth first but
instead a sort of FIFO order requried by the ls standard.</p>

<a name="toys">
<h1><a href="#toys">Directory toys/</a></h1>

<p>This directory contains command implementations. Each command is a single
self-contained file. Adding a new command involves adding a single
file, and removing a command involves removing that file. Commands use
shared infrastructure from the lib/ and generated/ directories.</p>

<p>Currently there are five subdirectories under "toys/" containing "posix"
commands described in POSIX-2008, "lsb" commands described in the Linux
Standard Base 4.1, "other" commands not described by either standard,
"pending" commands awaiting cleanup (which default to "n" in menuconfig
because they don't necessarily work right yet), and "example" code showing
how toybox infrastructure works and providing template/skeleton files to
start new commands.</p>

<p>The only difference directory location makes is which menu the command
shows up in during "make menuconfig", the directories are otherwise identical.
Note that the commands exist within a single namespace at runtime, so you can't
have the same command in multiple subdirectories. (The build tries to fail
informatively when you do that.)</p>

<p>There is one more sub-menus in "make menuconfig" containing global
configuration options for toybox. This menu is defined in the top level
Config.in.</p>

<p>See <a href="#adding">adding a new command</a> for details on the
layout of a command file.</p>

<a name="scripts">
<h2>Directory scripts/</h2>

<p>Build infrastructure. The makefile calls scripts/make.sh for "make"
and scripts/install.sh for "make install".</p>

<p>There's also a test suite, "make test" calls make/test.sh, which runs all
the tests in make/test/*. You can run individual tests via
"scripts/test.sh command", or "TEST_HOST=1 scripts/test.sh command" to run
that test against the host implementation instead of the toybox one.</p>

<h3>scripts/cfg2files.sh</h3>

<p>Run .config through this filter to get a list of enabled commands, which
is turned into a list of files in toys via a sed invocation in the top level
Makefile.
</p>

<h2>Directory kconfig/</h2>

<p>Menuconfig infrastructure copied from the Linux kernel a long time ago
(version 2.6.16).  See the
Linux kernel's Documentation/kbuild/kconfig-language.txt</p>

<!-- todo

Better OLDTOY and multiple command explanation. From Config.in:

<p>A command with multiple names (or multiple similar commands implemented in
the same .c file) should have config symbols prefixed with the name of their
C file. I.E. config symbol prefixes are NEWTOY() names. If OLDTOY() names
have config symbols they must be options (symbols with an underscore and
suffix) to the NEWTOY() name. (See generated/toylist.h)</p>
-->

<!--#include file="footer.html" -->