Annotation of /trunk/mkinitrd-magellan/busybox/TODO
Parent Directory | Revision Log
Revision 816 -
(hide annotations)
(download)
Fri Apr 24 18:33:46 2009 UTC (15 years, 5 months ago) by niro
File size: 13305 byte(s)
Fri Apr 24 18:33:46 2009 UTC (15 years, 5 months ago) by niro
File size: 13305 byte(s)
-updated to busybox-1.13.4
1 | niro | 532 | Busybox TODO |
2 | |||
3 | Stuff that needs to be done. This is organized by who plans to get around to | ||
4 | doing it eventually, but that doesn't mean they "own" the item. If you want to | ||
5 | do one of these bounce an email off the person it's listed under to see if they | ||
6 | have any suggestions how they plan to go about it, and to minimize conflicts | ||
7 | between your work and theirs. But otherwise, all of these are fair game. | ||
8 | |||
9 | niro | 816 | Rob Landley suggested these: |
10 | niro | 532 | Add a libbb/platform.c |
11 | Implement fdprintf() for platforms that haven't got one. | ||
12 | Implement bb_realpath() that can handle NULL on non-glibc. | ||
13 | Cleanup bb_asprintf() | ||
14 | |||
15 | Remove obsolete _() wrapper crud for internationalization we don't do. | ||
16 | Figure out where we need utf8 support, and add it. | ||
17 | |||
18 | sh | ||
19 | niro | 816 | The command shell situation is a big mess. We have three different |
20 | niro | 532 | shells that don't really share any code, and the "standalone shell" doesn't |
21 | work all that well (especially not in a chroot environment), due to apps not | ||
22 | niro | 816 | being reentrant. |
23 | lash is phased out. hush can be configured down to be nearly as small, | ||
24 | but less buggy :) | ||
25 | niro | 532 | init |
26 | General cleanup (should use ENABLE_FEATURE_INIT_SYSLOG and ENABLE_FEATURE_INIT_DEBUG). | ||
27 | Do a SUSv3 audit | ||
28 | Look at the full Single Unix Specification version 3 (available online at | ||
29 | "http://www.opengroup.org/onlinepubs/009695399/nfindex.html") and | ||
30 | figure out which of our apps are compliant, and what we're missing that | ||
31 | we might actually care about. | ||
32 | |||
33 | Even better would be some kind of automated compliance test harness that | ||
34 | exercises each command line option and the various corner cases. | ||
35 | Internationalization | ||
36 | How much internationalization should we do? | ||
37 | |||
38 | The low hanging fruit is UTF-8 character set support. We should do this. | ||
39 | (Vodz pointed out the shell's cmdedit as needing work here. What else?) | ||
40 | |||
41 | We also have lots of hardwired english text messages. Consolidating this | ||
42 | into some kind of message table not only makes translation easier, but | ||
43 | also allows us to consolidate redundant (or close) strings. | ||
44 | |||
45 | We probably don't want to be bloated with locale support. (Not unless we | ||
46 | can cleanly export it from our underlying C library without having to | ||
47 | concern ourselves with it directly. Perhaps a few specific things like a | ||
48 | config option for "date" are low hanging fruit here?) | ||
49 | |||
50 | What level should things happen at? How much do we care about | ||
51 | internationalizing the text console when X11 and xterms are so much better | ||
52 | at it? (There's some infrastructure here we don't implement: The | ||
53 | "unicode_start" and "unicode_stop" shell scripts need "vt-is-UTF8" and a | ||
54 | --unicode option to loadkeys. That implies a real loadkeys/dumpkeys | ||
55 | implementation to replace loadkmap/dumpkmap. Plus messing with console font | ||
56 | loading. Is it worth it, or do we just say "use X"?) | ||
57 | |||
58 | Individual compilation of applets. | ||
59 | It would be nice if busybox had the option to compile to individual applets, | ||
60 | for people who want an alternate implementation less bloated than the gnu | ||
61 | utils (or simply with less political baggage), but without it being one big | ||
62 | executable. | ||
63 | |||
64 | Turning libbb into a real dll is another possibility, especially if libbb | ||
65 | could export some of the other library interfaces we've already more or less | ||
66 | got the code for (like zlib). | ||
67 | buildroot - Make a "dogfood" option | ||
68 | Busybox 1.1 will be capable of replacing most gnu packages for real world | ||
69 | use, such as developing software or in a live CD. It needs wider testing. | ||
70 | |||
71 | Busybox should now be able to replace bzip2, coreutils, e2fsprogs, file, | ||
72 | findutils, gawk, grep, inetutils, less, modutils, net-tools, patch, procps, | ||
73 | sed, shadow, sysklogd, sysvinit, tar, util-linux, and vim. The resulting | ||
74 | system should be self-hosting (I.E. able to rebuild itself from source | ||
75 | code). This means it would need (at least) binutils, gcc, and make, or | ||
76 | equivalents. | ||
77 | |||
78 | It would be a good "eating our own dogfood" test if buildroot had the option | ||
79 | of using a "make allyesconfig" busybox instead of the all of the above | ||
80 | packages. Anything that's wrong with the resulting system, we can fix. (It | ||
81 | would be nice to be able to upgrade busybox to be able to replace bash and | ||
82 | diffutils as well, but we're not there yet.) | ||
83 | |||
84 | One example of an existing system that does this already is Firmware Linux: | ||
85 | http://www.landley.net/code/firmware | ||
86 | initramfs | ||
87 | Busybox should have a sample initramfs build script. This depends on | ||
88 | bbsh, mdev, and switch_root. | ||
89 | mkdep | ||
90 | Write a mkdep that doesn't segfault if there's a directory it doesn't | ||
91 | have permission to read, isn't based on manually editing the output of | ||
92 | lexx and yacc, doesn't make such a mess under include/config, etc. | ||
93 | Group globals into unions of structures. | ||
94 | Go through and turn all the global and static variables into structures, | ||
95 | and have all those structures be in a big union shared between processes, | ||
96 | so busybox uses less bss. (This is a big win on nommu machines.) See | ||
97 | sed.c and mdev.c for examples. | ||
98 | Go through bugs.busybox.net and close out all of that somehow. | ||
99 | This one's open to everybody, but I'll wind up doing it... | ||
100 | |||
101 | |||
102 | niro | 816 | Bernhard Reutner-Fischer <busybox@busybox.net> suggests to look at these: |
103 | niro | 532 | New debug options: |
104 | -Wlarger-than-127 | ||
105 | Cleanup any big users | ||
106 | -Wunused-parameter | ||
107 | Facilitate applet PROTOTYPES to provide means for having applets that | ||
108 | do a) not take any arguments b) need only one of argc or argv c) need | ||
109 | both argc and argv. All of these three options should go for the most | ||
110 | feature complete denominator. | ||
111 | Collate BUFSIZ IOBUF_SIZE MY_BUF_SIZE PIPE_PROGRESS_SIZE BUFSIZE PIPESIZE | ||
112 | make bb_common_bufsiz1 configurable, size wise. | ||
113 | make pipesize configurable, size wise. | ||
114 | Use bb_common_bufsiz1 throughout applets! | ||
115 | |||
116 | As yet unclaimed: | ||
117 | |||
118 | ---- | ||
119 | diff | ||
120 | Make sure we handle empty files properly: | ||
121 | From the patch man page: | ||
122 | |||
123 | you can remove a file by sending out a context diff that compares | ||
124 | the file to be deleted with an empty file dated the Epoch. The | ||
125 | file will be removed unless patch is conforming to POSIX and the | ||
126 | -E or --remove-empty-files option is not given. | ||
127 | --- | ||
128 | patch | ||
129 | Should have simple fuzz factor support to apply patches at an offset which | ||
130 | shouldn't take up too much space. | ||
131 | |||
132 | And while we're at it, a new patch filename quoting format is apparently | ||
133 | coming soon: http://marc.theaimsgroup.com/?l=git&m=112927316408690&w=2 | ||
134 | --- | ||
135 | man | ||
136 | It would be nice to have a man command. Not one that handles troff or | ||
137 | anything, just one that can handle preformatted ascii man pages, possibly | ||
138 | compressed. This could probably be a script in the extras directory that | ||
139 | calls cat/zcat/bzcat | less | ||
140 | |||
141 | (How doclifter might work into this is anybody's guess.) | ||
142 | --- | ||
143 | ar | ||
144 | niro | 816 | Write support! |
145 | --- | ||
146 | niro | 532 | stty / catv |
147 | stty's visible() function and catv's guts are identical. Merge them into | ||
148 | an appropriate libbb function. | ||
149 | niro | 816 | --- |
150 | struct suffix_mult | ||
151 | Several duplicate users of: grep -r "1024\*1024" * -B2 -A1 | ||
152 | Merge to a single size_suffixes[] in libbb. | ||
153 | Users: head tail od_bloaty hexdump and (partially as it wouldn't hurt) svlogd | ||
154 | --- | ||
155 | tail | ||
156 | ./busybox tail -f foo.c~ TODO | ||
157 | should not print fmt=header_fmt for subsequent date >> TODO; i.e. only | ||
158 | fmt+ if another (not the current) file did change | ||
159 | niro | 532 | |
160 | Architectural issues: | ||
161 | |||
162 | bb_close() with fsync() | ||
163 | We should have a bb_close() in place of normal close, with a CONFIG_ option | ||
164 | to not just check the return value of close() for an error, but fsync(). | ||
165 | Close can't reliably report anything useful because if write() accepted the | ||
166 | data then it either went out to the network or it's in cache or a pipe | ||
167 | buffer. Either way, there's no guarantee it'll make it to its final | ||
168 | destination before close() gets called, so there's no guarantee that any | ||
169 | error will be reported. | ||
170 | |||
171 | You need to call fsync() if you care about errors that occur after write(), | ||
172 | but that can have a big performance impact. So make it a config option. | ||
173 | --- | ||
174 | Unify archivers | ||
175 | Lots of archivers have the same general infrastructure. The directory | ||
176 | traversal code should be factored out, and the guts of each archiver could | ||
177 | be some setup code and a series of callbacks for "add this file", | ||
178 | "add this directory", "add this symlink" and so on. | ||
179 | |||
180 | This could clean up tar and zip, and make it cheaper to add cpio and ar | ||
181 | write support, and possibly even cheaply add things like mkisofs or | ||
182 | mksquashfs someday, if they become relevant. | ||
183 | --- | ||
184 | Text buffer support. | ||
185 | Several existing applets (sort, vi, less...) read | ||
186 | a whole file into memory and act on it. There might be an opportunity | ||
187 | for shared code in there that could be moved into libbb... | ||
188 | --- | ||
189 | Memory Allocation | ||
190 | We have a CONFIG_BUFFER mechanism that lets us select whether to do memory | ||
191 | allocation on the stack or the heap. Unfortunately, we're not using it much. | ||
192 | We need to audit our memory allocations and turn a lot of malloc/free calls | ||
193 | into RESERVE_CONFIG_BUFFER/RELEASE_CONFIG_BUFFER. | ||
194 | For a start, see e.g. make EXTRA_CFLAGS=-Wlarger-than-64 | ||
195 | |||
196 | And while we're at it, many of the CONFIG_FEATURE_CLEAN_UP #ifdefs will be | ||
197 | optimized out by the compiler in the stack allocation case (since there's no | ||
198 | free for an alloca()), and this means that various cleanup loops that just | ||
199 | call free might also be optimized out by the compiler if written right, so | ||
200 | we can yank those #ifdefs too, and generally clean up the code. | ||
201 | --- | ||
202 | Switch CONFIG_SYMBOLS to ENABLE_SYMBOLS | ||
203 | |||
204 | In busybox 1.0 and earlier, configuration was done by CONFIG_SYMBOLS | ||
205 | that were either defined or undefined to indicate whether the symbol was | ||
206 | selected in the .config file. They were used with #ifdefs, ala: | ||
207 | |||
208 | #ifdef CONFIG_SYMBOL | ||
209 | if (other_test) { | ||
210 | do_code(); | ||
211 | } | ||
212 | #endif | ||
213 | |||
214 | In 1.1, we have new ENABLE_SYMBOLS which are always defined (as 0 or 1), | ||
215 | meaning you can still use them for preprocessor tests by replacing | ||
216 | "#ifdef CONFIG_SYMBOL" with "#if ENABLE_SYMBOL". But more importantly, we | ||
217 | can use them as a true or false test in normal C code: | ||
218 | |||
219 | if (ENABLE_SYMBOL && other_test) { | ||
220 | do_code(); | ||
221 | } | ||
222 | |||
223 | (Optimizing away if() statements that resolve to a constant value | ||
224 | is known as "dead code elimination", an optimization so old and simple that | ||
225 | Turbo Pascal for DOS did it twenty years ago. Even modern mini-compilers | ||
226 | like the Tiny C Compiler (tcc) and the Small Device C Compiler (SDCC) | ||
227 | perform dead code elimination.) | ||
228 | |||
229 | Right now, busybox.h is #including both "config.h" (defining the | ||
230 | CONFIG_SYMBOLS) and "bb_config.h" (defining the ENABLE_SYMBOLS). At some | ||
231 | point in the future, it would be nice to wean ourselves off of the | ||
232 | CONFIG versions. (Among other things, some defective build environments | ||
233 | leak the Linux kernel's CONFIG_SYMBOLS into the system's standard #include | ||
234 | files. We've experienced collisions before.) | ||
235 | --- | ||
236 | FEATURE_CLEAN_UP | ||
237 | This is more an unresolved issue than a to-do item. More thought is needed. | ||
238 | |||
239 | Normally we rely on exit() to free memory, close files, and unmap segments | ||
240 | for us. This makes most calls to free(), close(), and unmap() optional in | ||
241 | busybox applets that don't intend to run for very long, and optional stuff | ||
242 | can be omitted to save size. | ||
243 | |||
244 | The idea was raised that we could simulate fork/exit with setjmp/longjmp | ||
245 | for _really_ brainless embedded systems, or speed up the standalone shell | ||
246 | by not forking. Doing so would require a reliable FEATURE_CLEAN_UP. | ||
247 | Unfortunately, this isn't as easy as it sounds. | ||
248 | |||
249 | The problem is, lots of things exit(), sometimes unexpectedly (xmalloc()) | ||
250 | and sometimes reliably (bb_perror_msg_and_die() or show_usage()). This | ||
251 | jumps out of the normal flow control and bypasses any cleanup code we | ||
252 | put at the end of our applets. | ||
253 | |||
254 | It's possible to add hooks to libbb functions like xmalloc() and xopen() | ||
255 | to add their entries to a linked list, which could be traversed and | ||
256 | freed/closed automatically. (This would need to be able to free just the | ||
257 | entries after a checkpoint to be usable for a forkless standalone shell. | ||
258 | You don't want to free the shell's own resources.) | ||
259 | |||
260 | Right now, FEATURE_CLEAN_UP is more or less a debugging aid, to make things | ||
261 | like valgrind happy. It's also documentation of _what_ we're trusting | ||
262 | exit() to clean up for us. But new infrastructure to auto-free stuff would | ||
263 | render the existing FEATURE_CLEAN_UP code redundant. | ||
264 | |||
265 | For right now, exit() handles it just fine. | ||
266 | |||
267 | |||
268 | |||
269 | Minor stuff: | ||
270 | watchdog.c could autodetect the timer duration via: | ||
271 | if(!ioctl (fd, WDIOC_GETTIMEOUT, &tmo)) timer_duration = 1 + (tmo / 2); | ||
272 | Unfortunately, that needs linux/watchdog.h and that contains unfiltered | ||
273 | kernel types on some distros, which breaks the build. | ||
274 | --- | ||
275 | use bb_error_msg where appropriate: See | ||
276 | egrep "(printf.*\([[:space:]]*(stderr|2)|[^_]write.*\([[:space:]]*(stderr|2))" | ||
277 | --- | ||
278 | use bb_perror_msg where appropriate: See | ||
279 | egrep "[^_]perror" | ||
280 | --- | ||
281 | possible code duplication ingroup() and is_a_group_member() | ||
282 | --- | ||
283 | Move __get_hz() to a better place and (re)use it in route.c, ash.c, msh.c | ||
284 | --- | ||
285 | niro | 816 | See grep -r strtod |
286 | Alot of duplication that wants cleanup. | ||
287 | --- | ||
288 | in_ether duplicated in network/{interface,ifconfig}.c | ||
289 | --- | ||
290 | niro | 532 | |
291 | |||
292 | Code cleanup: | ||
293 | |||
294 | Replace deprecated functions. | ||
295 | |||
296 | --- | ||
297 | vdprintf() -> similar sized functionality | ||
298 | --- |