source: MondoRescue/branches/stable/mindi-busybox/TODO @ 821

Last change on this file since 821 was 821, checked in by Bruno Cornec, 14 years ago

Addition of busybox 1.2.1 as a mindi-busybox new package
This should avoid delivering binary files in mindi not built there (Fedora and Debian are quite serious about that)

File size: 13.4 KB
1Busybox TODO
3Stuff that needs to be done.  This is organized by who plans to get around to
4doing it eventually, but that doesn't mean they "own" the item.  If you want to
5do one of these bounce an email off the person it's listed under to see if they
6have any suggestions how they plan to go about it, and to minimize conflicts
7between your work and theirs.  But otherwise, all of these are fair game.
9Rob Landley <>:
10  Add BB_NOMMU to platform.h and migrate __uClinux__ tests to that.
11    #if defined __UCLIBC__ && !defined __ARCH_USE_MMU__
12  Add a libbb/platform.c
13    Implement fdprintf() for platforms that haven't got one.
14    Implement bb_realpath() that can handle NULL on non-glibc.
15    Cleanup bb_asprintf()
17  Migrate calloc() and bb_calloc() occurrences to bb_xzalloc().
18  Remove obsolete _() wrapper crud for internationalization we don't do.
19  Figure out where we need utf8 support, and add it.
21  sh
22    The command shell situation is a big mess.  We have three or four different
23    shells that don't really share any code, and the "standalone shell" doesn't
24    work all that well (especially not in a chroot environment), due to apps not
25    being reentrant.  I'm writing a new shell (bbsh) to unify the various
26    shells and configurably add the minimal set of bash features people
27    actually use.  The hardest part is it has to configure down as small as
28    lash while providing lash's features.  The rest is easy in comparison.
29  bzip2
30    Compression-side support.
31  init
32    General cleanup (should use ENABLE_FEATURE_INIT_SYSLOG and ENABLE_FEATURE_INIT_DEBUG).
33  depmod
34    busybox lacks a way to update module deps when running from firmware without the
35    use of the (perl is to bloated for most embedded setups) and or orig
36    modutils. The orig depmod is rather pointless to have to add to a firmware image
37    in when we already have a insmod/rmmod and friends.
38  Unify base64 handling.
39    There's base64 encoding and decoding going on in:
40      networking/wget.c:base64enc()
41      coreutils/uudecode.c:read_base64()
42      coreutils/uuencode.c:tbl_base64[]
43      networking/httpd.c:decodeBase64()
44    And probably elsewhere.  That needs to be unified into libbb functions.
45  Do a SUSv3 audit
46    Look at the full Single Unix Specification version 3 (available online at
47    "") and
48    figure out which of our apps are compliant, and what we're missing that
49    we might actually care about.
51    Even better would be some kind of automated compliance test harness that
52    exercises each command line option and the various corner cases.
53  Internationalization
54    How much internationalization should we do?
56    The low hanging fruit is UTF-8 character set support.  We should do this.
57    (Vodz pointed out the shell's cmdedit as needing work here.  What else?)
59    We also have lots of hardwired english text messages.  Consolidating this
60    into some kind of message table not only makes translation easier, but
61    also allows us to consolidate redundant (or close) strings.
63    We probably don't want to be bloated with locale support.  (Not unless we
64    can cleanly export it from our underlying C library without having to
65    concern ourselves with it directly.  Perhaps a few specific things like a
66    config option for "date" are low hanging fruit here?)
68    What level should things happen at?  How much do we care about
69    internationalizing the text console when X11 and xterms are so much better
70    at it?  (There's some infrastructure here we don't implement: The
71    "unicode_start" and "unicode_stop" shell scripts need "vt-is-UTF8" and a
72    --unicode option to loadkeys.  That implies a real loadkeys/dumpkeys
73    implementation to replace loadkmap/dumpkmap.  Plus messing with console font
74    loading.  Is it worth it, or do we just say "use X"?)
76  Individual compilation of applets.
77    It would be nice if busybox had the option to compile to individual applets,
78    for people who want an alternate implementation less bloated than the gnu
79    utils (or simply with less political baggage), but without it being one big
80    executable.
82    Turning libbb into a real dll is another possibility, especially if libbb
83    could export some of the other library interfaces we've already more or less
84    got the code for (like zlib).
85  buildroot - Make a "dogfood" option
86    Busybox 1.1 will be capable of replacing most gnu packages for real world
87    use, such as developing software or in a live CD.  It needs wider testing.
89    Busybox should now be able to replace bzip2, coreutils, e2fsprogs, file,
90    findutils, gawk, grep, inetutils, less, modutils, net-tools, patch, procps,
91    sed, shadow, sysklogd, sysvinit, tar, util-linux, and vim.  The resulting
92    system should be self-hosting (I.E. able to rebuild itself from source
93    code).  This means it would need (at least) binutils, gcc, and make, or
94    equivalents.
96    It would be a good "eating our own dogfood" test if buildroot had the option
97    of using a "make allyesconfig" busybox instead of the all of the above
98    packages.  Anything that's wrong with the resulting system, we can fix.  (It
99    would be nice to be able to upgrade busybox to be able to replace bash and
100    diffutils as well, but we're not there yet.)
102    One example of an existing system that does this already is Firmware Linux:
104  initramfs
105    Busybox should have a sample initramfs build script.  This depends on
106    bbsh, mdev, and switch_root.
107  mkdep
108    Write a mkdep that doesn't segfault if there's a directory it doesn't
109    have permission to read, isn't based on manually editing the output of
110    lexx and yacc, doesn't make such a mess under include/config, etc.
111  Group globals into unions of structures.
112    Go through and turn all the global and static variables into structures,
113    and have all those structures be in a big union shared between processes,
114    so busybox uses less bss.  (This is a big win on nommu machines.)  See
115    sed.c and mdev.c for examples.
116  Go through and close out all of that somehow.
117    This one's open to everybody, but I'll wind up doing it...
120Bernhard Fischer <>:
121  Makefile stuff:
122    make -j is broken, -j1 is forced atm
123  New debug options:
124    -Wlarger-than-127
126    Use bb_common_bufsiz1?
128As yet unclaimed:
132  doesn't understand (), lots of susv3 stuff.
135  Make sure we handle empty files properly:
136    From the patch man page:
138   you can remove a file by sending out a context diff that compares
139   the file to be deleted with an empty file dated the Epoch.  The
140   file will be removed unless patch is conforming to POSIX and the
141   -E or --remove-empty-files option is not given.
144  Should have simple fuzz factor support to apply patches at an offset which
145  shouldn't take up too much space.
147  And while we're at it, a new patch filename quoting format is apparently
148  coming soon:
151  It would be nice to have a man command.  Not one that handles troff or
152  anything, just one that can handle preformatted ascii man pages, possibly
153  compressed.  This could probably be a script in the extras directory that
154  calls cat/zcat/bzcat | less
156  (How doclifter might work into this is anybody's guess.)
159  Write support?
164Architectural issues:
166bb_close() with fsync()
167  We should have a bb_close() in place of normal close, with a CONFIG_ option
168  to not just check the return value of close() for an error, but fsync().
169  Close can't reliably report anything useful because if write() accepted the
170  data then it either went out to the network or it's in cache or a pipe
171  buffer.  Either way, there's no guarantee it'll make it to its final
172  destination before close() gets called, so there's no guarantee that any
173  error will be reported.
175  You need to call fsync() if you care about errors that occur after write(),
176  but that can have a big performance impact.  So make it a config option.
178Unify archivers
179  Lots of archivers have the same general infrastructure.  The directory
180  traversal code should be factored out, and the guts of each archiver could
181  be some setup code and a series of callbacks for "add this file",
182  "add this directory", "add this symlink" and so on.
184  This could clean up tar and zip, and make it cheaper to add cpio and ar
185  write support, and possibly even cheaply add things like mkisofs or
186  mksquashfs someday, if they become relevant.
188Text buffer support.
189  Several existing applets (sort, vi, less...) read
190  a whole file into memory and act on it.  There might be an opportunity
191  for shared code in there that could be moved into libbb...
193Memory Allocation
194  We have a CONFIG_BUFFER mechanism that lets us select whether to do memory
195  allocation on the stack or the heap.  Unfortunately, we're not using it much.
196  We need to audit our memory allocations and turn a lot of malloc/free calls
198  For a start, see e.g. make EXTRA_CFLAGS=-Wlarger-than-64
200  And while we're at it, many of the CONFIG_FEATURE_CLEAN_UP #ifdefs will be
201  optimized out by the compiler in the stack allocation case (since there's no
202  free for an alloca()), and this means that various cleanup loops that just
203  call free might also be optimized out by the compiler if written right, so
204  we can yank those #ifdefs too, and generally clean up the code.
208  In busybox 1.0 and earlier, configuration was done by CONFIG_SYMBOLS
209  that were either defined or undefined to indicate whether the symbol was
210  selected in the .config file.  They were used with #ifdefs, ala:
212    #ifdef CONFIG_SYMBOL
213      if (other_test) {
214        do_code();
215      }
216    #endif
218  In 1.1, we have new ENABLE_SYMBOLS which are always defined (as 0 or 1),
219  meaning you can still use them for preprocessor tests by replacing
220  "#ifdef CONFIG_SYMBOL" with "#if ENABLE_SYMBOL".  But more importantly, we
221  can use them as a true or false test in normal C code:
223    if (ENABLE_SYMBOL && other_test) {
224      do_code();
225    }
227  (Optimizing away if() statements that resolve to a constant value
228  is known as "dead code elimination", an optimization so old and simple that
229  Turbo Pascal for DOS did it twenty years ago.  Even modern mini-compilers
230  like the Tiny C Compiler (tcc) and the Small Device C Compiler (SDCC)
231  perform dead code elimination.)
233  Right now, busybox.h is #including both "config.h" (defining the
234  CONFIG_SYMBOLS) and "bb_config.h" (defining the ENABLE_SYMBOLS).  At some
235  point in the future, it would be nice to wean ourselves off of the
236  CONFIG versions.  (Among other things, some defective build environments
237  leak the Linux kernel's CONFIG_SYMBOLS into the system's standard #include
238  files.  We've experienced collisions before.)
241  This is more an unresolved issue than a to-do item.  More thought is needed.
243  Normally we rely on exit() to free memory, close files, and unmap segments
244  for us.  This makes most calls to free(), close(), and unmap() optional in
245  busybox applets that don't intend to run for very long, and optional stuff
246  can be omitted to save size.
248  The idea was raised that we could simulate fork/exit with setjmp/longjmp
249  for _really_ brainless embedded systems, or speed up the standalone shell
250  by not forking.  Doing so would require a reliable FEATURE_CLEAN_UP.
251  Unfortunately, this isn't as easy as it sounds.
253  The problem is, lots of things exit(), sometimes unexpectedly (xmalloc())
254  and sometimes reliably (bb_perror_msg_and_die() or show_usage()).  This
255  jumps out of the normal flow control and bypasses any cleanup code we
256  put at the end of our applets.
258  It's possible to add hooks to libbb functions like xmalloc() and bb_xopen()
259  to add their entries to a linked list, which could be traversed and
260  freed/closed automatically.  (This would need to be able to free just the
261  entries after a checkpoint to be usable for a forkless standalone shell.
262  You don't want to free the shell's own resources.)
264  Right now, FEATURE_CLEAN_UP is more or less a debugging aid, to make things
265  like valgrind happy.  It's also documentation of _what_ we're trusting
266  exit() to clean up for us.  But new infrastructure to auto-free stuff would
267  render the existing FEATURE_CLEAN_UP code redundant.
269  For right now, exit() handles it just fine.
273Minor stuff:
274  watchdog.c could autodetect the timer duration via:
275    if(!ioctl (fd, WDIOC_GETTIMEOUT, &tmo)) timer_duration = 1 + (tmo / 2);
276  Unfortunately, that needs linux/watchdog.h and that contains unfiltered
277  kernel types on some distros, which breaks the build.
279  use bb_error_msg where appropriate: See
280  egrep "(printf.*\([[:space:]]*(stderr|2)|[^_]write.*\([[:space:]]*(stderr|2))"
282  use bb_perror_msg where appropriate: See
283  egrep "[^_]perror"
285  Remove superfluous fmt occurances: e.g.
286  fprintf(stderr, "%s: %s not found\n", "unalias", *argptr);
287  -> fprintf(stderr, "unalias: %s not found\n", *argptr);
289  possible code duplication ingroup() and is_a_group_member()
291  unify itoa: netstat.c, hush.c, lash.c, msh.c
292  Put one single, robust version into e.g. safe_strtol.c
296Code cleanup:
298Replace deprecated functions.
300bzero() -> memset()
302sigblock(), siggetmask(), sigsetmask(), sigmask() -> sigprocmask et al
304vdprintf() -> similar sized functionality
Note: See TracBrowser for help on using the repository browser.