ndk-busybox/TODO

Busybox TODO

Stuff that needs to be done.  This is organized by who plans to get around to
doing it eventually, but that doesn't mean they "own" the item.  If you want to
do one of these bounce an email off the person it's listed under to see if they
have any suggestions how they plan to go about it, and to minimize conflicts
between your work and theirs.  But otherwise, all of these are fair game.

Rob Landley <rob@landley.net>:
  Add BB_NOMMU to platform.h and migrate __uClinux__ tests to that.
    #if defined __UCLIBC__ && !defined __ARCH_USE_MMU__
  Add a libbb/platform.c
    Implement fdprintf() for platforms that haven't got one.
    Implement bb_realpath() that can handle NULL on non-glibc.
    Cleanup bb_asprintf()

  Migrate calloc() and bb_calloc() occurrences to bb_xzalloc().
  Remove obsolete _() wrapper crud for internationalization we don't do.
  Figure out where we need utf8 support, and add it.

  sh
    The command shell situation is a big mess.  We have three or four different
    shells that don't really share any code, and the "standalone shell" doesn't
    work all that well (especially not in a chroot environment), due to apps not
    being reentrant.  I'm writing a new shell (bbsh) to unify the various
    shells and configurably add the minimal set of bash features people
    actually use.  The hardest part is it has to configure down as small as
    lash while providing lash's features.  The rest is easy in comparison.
  bzip2
    Compression-side support.
  init
    General cleanup (should use ENABLE_FEATURE_INIT_SYSLOG and ENABLE_FEATURE_INIT_DEBUG).
  depmod
    busybox lacks a way to update module deps when running from firmware without the
    use of the depmod.pl (perl is to bloated for most embedded setups) and or orig
    modutils. The orig depmod is rather pointless to have to add to a firmware image
    in when we already have a insmod/rmmod and friends.
  Unify base64 handling.
    There's base64 encoding and decoding going on in:
      networking/wget.c:base64enc()
      coreutils/uudecode.c:read_base64()
      coreutils/uuencode.c:tbl_base64[]
      networking/httpd.c:decodeBase64()
    And probably elsewhere.  That needs to be unified into libbb functions.
  Do a SUSv3 audit
    Look at the full Single Unix Specification version 3 (available online at
    "http://www.opengroup.org/onlinepubs/009695399/nfindex.html") and
    figure out which of our apps are compliant, and what we're missing that
    we might actually care about.

    Even better would be some kind of automated compliance test harness that
    exercises each command line option and the various corner cases.
  Internationalization
    How much internationalization should we do?

    The low hanging fruit is UTF-8 character set support.  We should do this.
    (Vodz pointed out the shell's cmdedit as needing work here.  What else?)

    We also have lots of hardwired english text messages.  Consolidating this
    into some kind of message table not only makes translation easier, but
    also allows us to consolidate redundant (or close) strings.

    We probably don't want to be bloated with locale support.  (Not unless we
    can cleanly export it from our underlying C library without having to
    concern ourselves with it directly.  Perhaps a few specific things like a
    config option for "date" are low hanging fruit here?)

    What level should things happen at?  How much do we care about
    internationalizing the text console when X11 and xterms are so much better
    at it?  (There's some infrastructure here we don't implement: The
    "unicode_start" and "unicode_stop" shell scripts need "vt-is-UTF8" and a
    --unicode option to loadkeys.  That implies a real loadkeys/dumpkeys
    implementation to replace loadkmap/dumpkmap.  Plus messing with console font
    loading.  Is it worth it, or do we just say "use X"?)

  Individual compilation of applets.
    It would be nice if busybox had the option to compile to individual applets,
    for people who want an alternate implementation less bloated than the gnu
    utils (or simply with less political baggage), but without it being one big
    executable.

    Turning libbb into a real dll is another possibility, especially if libbb
    could export some of the other library interfaces we've already more or less
    got the code for (like zlib).
  buildroot - Make a "dogfood" option
    Busybox 1.1 will be capable of replacing most gnu packages for real world
    use, such as developing software or in a live CD.  It needs wider testing.

    Busybox should now be able to replace bzip2, coreutils, e2fsprogs, file,
    findutils, gawk, grep, inetutils, less, modutils, net-tools, patch, procps,
    sed, shadow, sysklogd, sysvinit, tar, util-linux, and vim.  The resulting
    system should be self-hosting (I.E. able to rebuild itself from source
    code).  This means it would need (at least) binutils, gcc, and make, or
    equivalents.

    It would be a good "eating our own dogfood" test if buildroot had the option
    of using a "make allyesconfig" busybox instead of the all of the above
    packages.  Anything that's wrong with the resulting system, we can fix.  (It
    would be nice to be able to upgrade busybox to be able to replace bash and
    diffutils as well, but we're not there yet.)

    One example of an existing system that does this already is Firmware Linux:
      http://www.landley.net/code/firmware
  initramfs
    Busybox should have a sample initramfs build script.  This depends on
    bbsh, mdev, and switch_root.
  mkdep
    Write a mkdep that doesn't segfault if there's a directory it doesn't
    have permission to read, isn't based on manually editing the output of
    lexx and yacc, doesn't make such a mess under include/config, etc.
  Group globals into unions of structures.
    Go through and turn all the global and static variables into structures,
    and have all those structures be in a big union shared between processes,
    so busybox uses less bss.  (This is a big win on nommu machines.)  See
    sed.c and mdev.c for examples.
  Go through bugs.busybox.net and close out all of that somehow.
    This one's open to everybody, but I'll wind up doing it...


Bernhard Fischer <busybox@busybox.net> suggests to look at these:
  Makefile stuff:
    make -j is broken. klibc make infrastructure (vda?)
  New debug options:
    -Wlarger-than-127
    Cleanup any big users
    -Wunused-parameter
    Facilitate applet PROTOTYPES to provide means for having applets that
    do a) not take any arguments b) need only one of argc or argv c) need
    both argc and argv. All of these three options should go for the most
    feature complete denominator.
  Collate BUFSIZ IOBUF_SIZE MY_BUF_SIZE PIPE_PROGRESS_SIZE BUFSIZE PIPESIZE
    make bb_common_bufsiz1 configurable, size wise.
    make pipesize configurable, size wise.
    Use bb_common_bufsiz1 throughout applets!

As yet unclaimed:

----
find
  doesn't understand (), lots of susv3 stuff.
----
diff
  Make sure we handle empty files properly:
    From the patch man page:

    you can remove a file by sending out a context diff that compares
    the file to be deleted with an empty file dated the Epoch.  The
    file will be removed unless patch is conforming to POSIX and the
    -E or --remove-empty-files option is not given.
---
patch
  Should have simple fuzz factor support to apply patches at an offset which
  shouldn't take up too much space.

  And while we're at it, a new patch filename quoting format is apparently
  coming soon:  http://marc.theaimsgroup.com/?l=git&m=112927316408690&w=2
---
man
  It would be nice to have a man command.  Not one that handles troff or
  anything, just one that can handle preformatted ascii man pages, possibly
  compressed.  This could probably be a script in the extras directory that
  calls cat/zcat/bzcat | less

  (How doclifter might work into this is anybody's guess.)
---
ar
  Write support?
---
crond
  turn FEATURE_DEBUG_OPT into ENABLE_FEATURE_CROND_DEBUG_OPT

Architectural issues:

bb_close() with fsync()
  We should have a bb_close() in place of normal close, with a CONFIG_ option
  to not just check the return value of close() for an error, but fsync().
  Close can't reliably report anything useful because if write() accepted the
  data then it either went out to the network or it's in cache or a pipe
  buffer.  Either way, there's no guarantee it'll make it to its final
  destination before close() gets called, so there's no guarantee that any
  error will be reported.

  You need to call fsync() if you care about errors that occur after write(),
  but that can have a big performance impact.  So make it a config option.
---
Unify archivers
  Lots of archivers have the same general infrastructure.  The directory
  traversal code should be factored out, and the guts of each archiver could
  be some setup code and a series of callbacks for "add this file",
  "add this directory", "add this symlink" and so on.

  This could clean up tar and zip, and make it cheaper to add cpio and ar
  write support, and possibly even cheaply add things like mkisofs or
  mksquashfs someday, if they become relevant.
---
Text buffer support.
  Several existing applets (sort, vi, less...) read
  a whole file into memory and act on it.  There might be an opportunity
  for shared code in there that could be moved into libbb...
---
Memory Allocation
  We have a CONFIG_BUFFER mechanism that lets us select whether to do memory
  allocation on the stack or the heap.  Unfortunately, we're not using it much.
  We need to audit our memory allocations and turn a lot of malloc/free calls
  into RESERVE_CONFIG_BUFFER/RELEASE_CONFIG_BUFFER.
  For a start, see e.g. make EXTRA_CFLAGS=-Wlarger-than-64

  And while we're at it, many of the CONFIG_FEATURE_CLEAN_UP #ifdefs will be
  optimized out by the compiler in the stack allocation case (since there's no
  free for an alloca()), and this means that various cleanup loops that just
  call free might also be optimized out by the compiler if written right, so
  we can yank those #ifdefs too, and generally clean up the code.
---
Switch CONFIG_SYMBOLS to ENABLE_SYMBOLS

  In busybox 1.0 and earlier, configuration was done by CONFIG_SYMBOLS
  that were either defined or undefined to indicate whether the symbol was
  selected in the .config file.  They were used with #ifdefs, ala:

    #ifdef CONFIG_SYMBOL
      if (other_test) {
        do_code();
      }
    #endif

  In 1.1, we have new ENABLE_SYMBOLS which are always defined (as 0 or 1),
  meaning you can still use them for preprocessor tests by replacing
  "#ifdef CONFIG_SYMBOL" with "#if ENABLE_SYMBOL".  But more importantly, we
  can use them as a true or false test in normal C code:

    if (ENABLE_SYMBOL && other_test) {
      do_code();
    }

  (Optimizing away if() statements that resolve to a constant value
  is known as "dead code elimination", an optimization so old and simple that
  Turbo Pascal for DOS did it twenty years ago.  Even modern mini-compilers
  like the Tiny C Compiler (tcc) and the Small Device C Compiler (SDCC)
  perform dead code elimination.)

  Right now, busybox.h is #including both "config.h" (defining the
  CONFIG_SYMBOLS) and "bb_config.h" (defining the ENABLE_SYMBOLS).  At some
  point in the future, it would be nice to wean ourselves off of the
  CONFIG versions.  (Among other things, some defective build environments
  leak the Linux kernel's CONFIG_SYMBOLS into the system's standard #include
  files.  We've experienced collisions before.)
---
FEATURE_CLEAN_UP
  This is more an unresolved issue than a to-do item.  More thought is needed.

  Normally we rely on exit() to free memory, close files, and unmap segments
  for us.  This makes most calls to free(), close(), and unmap() optional in
  busybox applets that don't intend to run for very long, and optional stuff
  can be omitted to save size.

  The idea was raised that we could simulate fork/exit with setjmp/longjmp
  for _really_ brainless embedded systems, or speed up the standalone shell
  by not forking.  Doing so would require a reliable FEATURE_CLEAN_UP.
  Unfortunately, this isn't as easy as it sounds.

  The problem is, lots of things exit(), sometimes unexpectedly (xmalloc())
  and sometimes reliably (bb_perror_msg_and_die() or show_usage()).  This
  jumps out of the normal flow control and bypasses any cleanup code we
  put at the end of our applets.

  It's possible to add hooks to libbb functions like xmalloc() and xopen()
  to add their entries to a linked list, which could be traversed and
  freed/closed automatically.  (This would need to be able to free just the
  entries after a checkpoint to be usable for a forkless standalone shell.
  You don't want to free the shell's own resources.)

  Right now, FEATURE_CLEAN_UP is more or less a debugging aid, to make things
  like valgrind happy.  It's also documentation of _what_ we're trusting
  exit() to clean up for us.  But new infrastructure to auto-free stuff would
  render the existing FEATURE_CLEAN_UP code redundant.

  For right now, exit() handles it just fine.



Minor stuff:
  watchdog.c could autodetect the timer duration via:
    if(!ioctl (fd, WDIOC_GETTIMEOUT, &tmo)) timer_duration = 1 + (tmo / 2);
  Unfortunately, that needs linux/watchdog.h and that contains unfiltered
  kernel types on some distros, which breaks the build.
---
  use bb_error_msg where appropriate: See
  egrep "(printf.*\([[:space:]]*(stderr|2)|[^_]write.*\([[:space:]]*(stderr|2))"
---
  use bb_perror_msg where appropriate: See
  egrep "[^_]perror"
---
  Remove superfluous fmt occurances: e.g.
  fprintf(stderr, "%s: %s not found\n", "unalias", *argptr);
  -> fprintf(stderr, "unalias: %s not found\n", *argptr);
---
  possible code duplication ingroup() and is_a_group_member()
---
  unify itoa: netstat.c, hush.c, lash.c, msh.c
  Put one single, robust version into e.g. safe_strtol.c
---
  Move __get_hz() to a better place and (re)use it in route.c, ash.c, msh.c
---


Code cleanup:

Replace deprecated functions.

bzero() -> memset()
---
sigblock(), siggetmask(), sigsetmask(), sigmask() -> sigprocmask et al
---
vdprintf() -> similar sized functionality
---