3 Stuff that needs to be done. All of this is fair game for 1.2.
6 doesn't understand (), lots of susv3 stuff.
9 The command shell situation is a big mess. We have three or four different
10 shells that don't really share any code, and the "standalone shell" doesn't
11 work all that well (especially not in a chroot environment), due to apps not
12 being reentrant. Unifying the various shells and figuring out a configurable
13 way of adding the minimal set of bash features a given script uses is a big
14 job, but it would be a big improvement.
16 Note: Rob Landley (rob@landley.net) is working on a new unified shell called
17 bbsh, but it's a low priority...
20 We should have a diff -u command. We have patch, we should have diff
21 (we only need to support unified diffs though).
23 Also, make sure we handle empty files properly:
24 From the patch man page:
26 you can remove a file by sending out a context diff that compares
27 the file to be deleted with an empty file dated the Epoch. The
28 file will be removed unless patch is conforming to POSIX and the
29 -E or --remove-empty-files option is not given.
32 Should have simple fuzz factor support to apply patches at an offset which
33 shouldn't take up too much space.
35 And while we're at it, a new patch filename quoting format is apparently
36 coming soon: http://marc.theaimsgroup.com/?l=git&m=112927316408690&w=2
39 It would be nice to have a man command. Not one that handles troff or
40 anything, just one that can handle preformatted ascii man pages, possibly
41 compressed. This could probably be a script in the extras directory that
42 calls cat/zcat/bzcat | less
44 (How doclifter might work into this is anybody's guess.)
47 Compression-side support.
60 bb_close() with fsync()
61 We should have a bb_close() in place of normal close, with a CONFIG_ option
62 to not just check the return value of close() for an error, but fsync().
63 Close can't reliably report anything useful because if write() accepted the
64 data then it either went out to the network or it's in cache or a pipe
65 buffer. Either way, there's no guarantee it'll make it to its final
66 destination before close() gets called, so there's no guarantee that any
67 error will be reported.
69 You need to call fsync() if you care about errors that occur after write(),
70 but that can have a big performance impact. So make it a config option.
73 Look at the full Single Unix Specification version 3 (available online at
74 "http://www.opengroup.org/onlinepubs/009695399/nfindex.html") and
75 figure out which of our apps are compliant, and what we're missing that
76 we might actually care about.
78 Even better would be some kind of automated compliance test harness that
79 exercises each command line option and the various corner cases.
82 How much internationalization should we do?
84 The low hanging fruit is UTF-8 character set support. We should do this.
85 (Vodz pointed out the shell's cmdedit as needing work here. What else?)
87 We also have lots of hardwired english text messages. Consolidating this
88 into some kind of message table not only makes translation easier, but
89 also allows us to consolidate redundant (or close) strings.
91 We probably don't want to be bloated with locale support. (Not unless we can
92 cleanly export it from our underlying C library without having to concern
93 ourselves with it directly. Perhaps a few specific things like a config
94 option for "date" are low hanging fruit here?)
96 What level should things happen at? How much do we care about
97 internationalizing the text console when X11 and xterms are so much better
98 at it? (There's some infrastructure here we don't implement: The
99 "unicode_start" and "unicode_stop" shell scripts need "vt-is-UTF8" and a
100 --unicode option to loadkeys. That implies a real loadkeys/dumpkeys
101 implementation to replace loadkmap/dumpkmap. Plus messing with console font
102 loading. Is it worth it, or do we just say "use X"?)
105 Lots of archivers have the same general infrastructure. The directory
106 traversal code should be factored out, and the guts of each archiver could
107 be some setup code and a series of callbacks for "add this file",
108 "add this directory", "add this symlink" and so on.
110 This could clean up tar and zip, and make it cheaper to add cpio and ar
111 write support, and possibly even cheaply add things like mkisofs or
112 mksquashfs someday, if they become relevant.
115 Several existing applets (sort, vi, less...) read
116 a whole file into memory and act on it. There might be an opportunity
117 for shared code in there that could be moved into libbb...
119 Individual compilation of applets.
120 It would be nice if busybox had the option to compile to individual applets,
121 for people who want an alternate implementation less bloated than the gnu
122 utils (or simply with less political baggage), but without it being one big
125 Turning libbb into a real dll is another possibility, especially if libbb
126 could export some of the other library interfaces we've already more or less
127 got the code for (like zlib).
129 buildroot - Make a "dogfood" option
130 Busybox 1.1 will be capable of replacing most gnu packages for real world use,
131 such as developing software or in a live CD. It needs wider testing.
133 Busybox should now be able to replace bzip2, coreutils, e2fsprogs, file,
134 findutils, gawk, grep, inetutils, less, modutils, net-tools, patch, procps,
135 sed, shadow, sysklogd, sysvinit, tar, util-linux, and vim. The resulting
136 system should be self-hosting (I.E. able to rebuild itself from source code).
137 This means it would need (at least) binutils, gcc, and make, or equivalents.
139 It would be a good "eating our own dogfood" test if buildroot had the option
140 of using a "make allyesconfig" busybox instead of the all of the above
141 packages. Anything that's wrong with the resulting system, we can fix. (It
142 would be nice to be able to upgrade busybox to be able to replace bash and
143 diffutils as well, but we're not there yet.)
145 One example of an existing system that does this already is Firmware Linux:
146 http://www.landley.net/code/firmware
149 Busybox should have a sample initramfs build script. This depends on
150 involves bbsh, mdev, and switch_root.
153 We have a CONFIG_BUFFER mechanism that lets us select whether to do memory
154 allocation on the stack or the heap. Unfortunately, we're not using it much.
155 We need to audit our memory allocations and turn a lot of malloc/free calls
156 into RESERVE_CONFIG_BUFFER/RELEASE_CONFIG_BUFFER.
158 And while we're at it, many of the CONFIG_FEATURE_CLEAN_UP #ifdefs will be
159 optimized out by the compiler in the stack allocation case (since there's no
160 free for an alloca()), and this means that various cleanup loops that just
161 call free might also be optimized out by the compiler if written right, so
162 we can yank those #ifdefs too, and generally clean up the code.
164 Switch CONFIG_SYMBOLS to ENABLE_SYMBOLS
166 In busybox 1.0 and earlier, configuration was done by CONFIG_SYMBOLS
167 that were either defined or undefined to indicate whether the symbol was
168 selected in the .config file. They were used with #ifdefs, ala:
176 In 1.1, we have new ENABLE_SYMBOLS which are always defined (as 0 or 1),
177 meaning you can still use them for preprocessor tests by replacing
178 "#ifdef CONFIG_SYMBOL" with "#if ENABLE_SYMBOL". But more importantly, we
179 can use them as a true or false test in normal C code:
181 if (ENABLE_SYMBOL && other_test) {
185 (Optimizing away if() statements that resolve to a constant value
186 is known as "dead code elimination", an optimization so old and simple that
187 Turbo Pascal for DOS did it twenty years ago. Even modern mini-compilers
188 like the Tiny C Compiler (tcc) and the Small Device C Compiler (SDCC)
189 perform dead code elimination.)
191 Right now, busybox.h is #including both "config.h" (defining the
192 CONFIG_SYMBOLS) and "bb_config.h" (defining the ENABLE_SYMBOLS). At some
193 point in the future, it would be nice to wean ourselves off of the
194 CONFIG versions. (Among other things, some defective build environments
195 leak the Linux kernel's CONFIG_SYMBOLS into the system's standard #include
196 files. We've experienced collisions before.)
199 This is more an unresolved issue than a to-do item. More thought is needed.
201 Normally we rely on exit() to free memory, close files, and unmap segments
202 for us. This makes most calls to free(), close(), and unmap() optional in
203 busybox applets that don't intend to run for very long, and optional stuff
204 can be omitted to save size.
206 The idea was raised that we could simulate fork/exit with setjmp/longjmp
207 for _really_ brainless embedded systems, or speed up the standalone shell
208 by not forking. Doing so would require a reliable FEATURE_CLEAN_UP.
209 Unfortunately, this isn't as easy as it sounds.
211 The problem is, lots of things exit(), sometimes unexpectedly (xmalloc())
212 and sometimes reliably (bb_perror_msg_and_die() or show_usage()). This
213 jumps out of the normal flow control and bypasses any cleanup code we
214 put at the end of our applets.
216 It's possible to add hooks to libbb functions like xmalloc() and bb_xopen()
217 to add their entries to a linked list, which could be traversed and
218 freed/closed automatically. (This would need to be able to free just the
219 entries after a checkpoint to be usable for a forkless standalone shell.
220 You don't want to free the shell's own resources.)
222 Right now, FEATURE_CLEAN_UP is more or less a debugging aid, to make things
223 like valgrind happy. It's also documentation of _what_ we're trusting
224 exit() to clean up for us. But new infrastructure to auto-free stuff would
225 render the existing FEATURE_CLEAN_UP code redundant.
227 For right now, exit() handles it just fine.
232 watchdog.c could autodetect the timer duration via:
233 if(!ioctl (fd, WDIOC_GETTIMEOUT, &tmo)) timer_duration = 1 + (tmo / 2);
234 Unfortunately, that needs linux/watchdog.h and that contains unfiltered
235 kernel types on some distros, which breaks the build.
240 Replace deprecated functions.
244 sigblock(), siggetmask(), sigsetmask(), sigmask() -> sigprocmask et al