00:12.24 | *** join/#uclibc thraxisp (n=thraxisp@ottawa-hs-206-191-33-40.d-ip.magma.ca) |
00:34.24 | *** join/#uclibc Serbitar (i=david@asguard.org.uk) |
01:24.23 | *** join/#uclibc blindvt (n=bf@M975P031.adsl.highway.telekom.at) |
01:52.26 | *** join/#uclibc hgb (n=hgb@tussi.moria.no) [NETSPLIT VICTIM] |
01:56.06 | *** join/#uclibc landley (n=landley@70.20.69.141) |
01:56.17 | landley | Things like bb_askpass() are way too obscure. |
01:56.51 | landley | It returns a pointer to a static buffer, so of course tito's password rewrite (which does simplify the code greatly) calls it once, calls it twice, and does a strcmp on the two pointers for "reenter your password" code. |
01:56.55 | landley | It's the _same_buffer_. |
01:57.01 | dalias | lol |
01:57.15 | dalias | then put 2 internal buffers |
01:57.24 | dalias | and a static int variable which switches between them each call :)) |
01:57.25 | landley | That would be "malloc". |
01:57.43 | landley | It's not entirely tito's fault, although I need to check all this stuff before actually checking it in. |
01:57.45 | dalias | sorry i was just trying to make it even more evil ;) |
01:57.53 | landley | It's a function that begs to be abused. |
01:58.00 | landley | Which is not a good thing to have anywhere near the password code. |
01:58.09 | dalias | i've actually made code like that once :P |
02:12.07 | *** join/#uclibc sjhill (n=sjhill@eth13.com-link.com) |
02:20.28 | dalias | btw landley did you hear earlier when i was talkingto psm? |
02:20.34 | dalias | ville is going to lgpl tre |
02:20.50 | dalias | (finally a good regex implementation!) |
02:23.29 | landley | Hmmm? |
02:23.36 | landley | Who is ville, and what is tre? |
02:24.01 | landley | (I wrote a regex implementation almost ten years ago. Kept meaning to rewrite the one in uClibc, but I was under the impression they'd already fixed it...) |
02:24.02 | dalias | TRE is the regex implementation that's smallest, fastest, and closest to posix compliance |
02:24.09 | landley | Ok. |
02:24.16 | landley | I take it ville wrote it? (Person or organization?) |
02:24.20 | dalias | person |
02:24.31 | dalias | gnu regex and rx are very broken and bloated |
02:25.30 | landley | Gnu anything is generally very bloated. Broken is optional. |
02:25.40 | landley | Dunno about rx. Is that what's in uClibc? |
02:25.40 | dalias | heh yeah no kidding |
02:26.02 | dalias | no, gnu regex (ancient version most likely) is in uclibc |
02:26.08 | landley | Suckage. |
02:26.22 | landley | Well, if the tre thing doesn't work out, let me know and I'll add it to my todo list. |
02:26.41 | dalias | i expect the current gnu version mallocs a copy of the whole string and preconverts it to wchar_t string... |
02:26.42 | landley | (It's one of those things you put off for a year and then do in a week...) |
02:26.45 | landley | I have many of those things. |
02:26.56 | landley | Ouch. |
02:27.09 | dalias | just like current version of gnu fnmatch does... |
02:27.48 | landley | I like ascii. |
02:27.51 | dalias | current gnu fnmatch seems to impose a length limit on the string (which violates posix, even tho in practice the strings are normally under PATH_MAX) so it can preconvert the whole thing to wchar_t |
02:28.19 | dalias | limiting to ascii is bad |
02:28.34 | landley | utf8 is ascii with a tailwind. |
02:28.36 | dalias | if you implement things sanely, supporting utf8 has essentially no overhead beyond ascii |
02:28.50 | landley | Being 8 bit clean is common sense. |
02:29.05 | dalias | 8bit-clean and utf8 are partly exclusive |
02:29.43 | landley | ? |
02:29.44 | dalias | fnmatch can do byte comparisons rather than character comparison most of the time, and fully support utf8 |
02:29.44 | dalias | however bracket expressions are lists of _characters_ not bytes |
02:29.44 | landley | How are 8 bit clean and and utf8 exclusive? |
02:29.55 | dalias | so for a bracket expression you need to process the expression and the string as characters |
02:30.04 | dalias | also for ? (in fnmatch) and . (in regex) |
02:30.09 | dalias | they match a character, not a byte |
02:30.13 | landley | Sure. |
02:30.35 | dalias | so what do you do if the next sequence in the string is not a valid character? :) |
02:30.46 | landley | Me? Treat it as bytes. |
02:31.05 | dalias | gnu behavior (lovely) is to fail to match, i expect |
02:31.17 | landley | There are three or four ways to express most utf8 characters anyway, I vaguely recall. |
02:31.18 | dalias | meaning that "rm *" cannot delete files with malformed utf8 in them, in utf8 locale... |
02:31.22 | dalias | nope |
02:31.24 | dalias | there is only one way |
02:31.45 | landley | dalias: the kernel ignores utf8 and treats everything as bytes. |
02:31.47 | dalias | if not utf8 would be very broken, like most other multibyte encodings |
02:31.50 | dalias | yes |
02:31.57 | dalias | as it probably should |
02:32.03 | landley | Being _able_ to get away with that is what made me decide utf8 was a good thing. |
02:32.13 | landley | Once again, we're back to the 99% solution. |
02:32.19 | dalias | my point was that i don't think "*" will match malformed sequences with gnu regex |
02:32.22 | dalias | errm |
02:32.24 | dalias | gnu fnmatch |
02:32.30 | landley | A few things (unfortunately sed and sort) may have to care. |
02:32.39 | landley | I say unfortunately because I handle those myself. :) |
02:32.51 | dalias | which is amusing because it makes it almost impossible for root to rm files with invalid sequences in them if a malicous user creates such a file :) |
02:33.12 | landley | I wonder what tab completion would do? |
02:33.17 | landley | Oh well, not going there today. |
02:33.19 | dalias | :) |
02:33.34 | landley | Right now, busybox tries very hard not to care about locale. |
02:33.43 | landley | At some point we may have to start, but for now? No. |
02:33.43 | dalias | in my implementation of fnmatch, * matches any number of bytes |
02:33.50 | dalias | and only . and [] care about the idea of characters |
02:33.56 | dalias | errm |
02:34.01 | dalias | ?, not . |
02:34.08 | dalias | keep confusing regex/fnmatch syntax :) |
02:34.17 | landley | I've implemented both. |
02:34.37 | landley | If I recall, [] is similar, ? is ., and * is .* |
02:34.44 | landley | Trying to remember if there were any other differences... |
02:34.45 | dalias | yeah |
02:34.58 | landley | Other than lack of {,} and () and so on. |
02:35.02 | dalias | well the difference is in the complexity of the languages |
02:35.38 | dalias | for fnmatch you always have exactly one position you're at in the pattern and in th estring (except for '*' but its easy to special-case it) |
02:35.55 | dalias | with regex, you have to keep track of multiple parallel possible states |
02:36.04 | landley | dalias: I know, I've done it. |
02:36.14 | dalias | :) |
02:36.20 | landley | It was ten years ago, but I implemented this crud back when I was working on OS/2 for IBM. |
02:36.25 | dalias | heh |
02:36.27 | dalias | i'm hoping to condense TRE to a tiny, compliant regex implementation |
02:36.37 | dalias | which we can use in both my libc and uclibc |
02:36.48 | landley | I vaguely remember I could start over from scratch and do it in a three day weekend. |
02:36.52 | landley | It wasn't _that_ hard of a problem. |
02:37.05 | landley | The hard part was making .*.*.*.*.*.*.*.* not blow up on you. :) |
02:37.12 | dalias | :) |
02:37.27 | landley | (Can current regex syntax specify greedy or miserly matches?) |
02:37.40 | landley | I remember I had a flag for that, don't remember how it was specified... |
02:37.41 | dalias | its always greedy |
02:37.44 | dalias | according to posix |
02:37.54 | landley | Is there a gnu extension for this? |
02:37.57 | dalias | if you dont want it to be, write your regex accordingly |
02:38.00 | dalias | not afaik |
02:38.06 | dalias | there's a perl extension for it |
02:38.08 | landley | I had both working, long ago... |
02:38.14 | dalias | but pcre is totally different from posix regex |
02:38.18 | landley | Is there a standard for perl regexes? |
02:38.23 | dalias | yes, perl :) |
02:38.38 | landley | Perl is not a standard. Perl is an implementation of larry wall's stream of consciousness. |
02:38.46 | dalias | exactly ;( |
02:38.49 | dalias | errm ;) |
02:38.56 | landley | Right, I'll worry about it later. |
02:39.06 | landley | Gotta beat passwd.c and obscure.c into shape. |
02:39.09 | dalias | anyway its just a matter of writing your regex appropriately |
02:39.11 | dalias | yeah |
02:39.32 | dalias | i should update by bb checkout and see what portability issues remain |
02:39.49 | landley | REG_NOTGREEDY |
02:39.58 | landley | dalias: rather a lot. |
02:40.03 | landley | I'm worrying about that for 1.1.2. |
02:40.22 | dalias | i'm going to try to make some new (hopefully as noncontroversial as possible) patches to clean up the dependence on glibc/linux headers, etc. |
02:40.40 | landley | There is now a platform.h, and localizing dependencies there is a good thing. |
02:40.53 | landley | You might want to talk to Shaun Jackman, he was getting busybox to work with newlib. |
02:41.55 | dalias | *nod* |
02:43.10 | dalias | i need to join the mailing list |
02:43.21 | landley | There's a web archive. :) |
02:43.25 | dalias | i know |
02:43.49 | dalias | but that doesnt work very well for participating |
02:43.56 | landley | Long ago anybody could post to the list, but then Erik got overwhelmed with spam. |
02:44.00 | dalias | yeah |
02:44.05 | landley | He was filtering it personally before then. Got to something like 1000 a day... |
02:44.11 | dalias | well i would not dream of using a list without subscribing |
02:44.15 | dalias | you'll break threading |
02:44.27 | landley | I read every message. :) |
02:44.33 | dalias | (i dont think the web archive has the message-id's) |
02:44.41 | landley | I seem to create half of 'em. :) |
02:44.47 | dalias | :) |
02:44.48 | landley | There's mbox files somewhere... |
02:44.52 | dalias | yeah |
02:44.52 | landley | Not that this is necessarily helpful. |
02:45.05 | dalias | but that would defeat the purpose -- might as well just subscibe |
02:50.02 | landley | bb_pwd.c is an abomination. |
02:50.34 | landley | Gluing a half-dozen .c files together (so that even their #includes are inside #ifdefs) is not an improvement over halfing a half-dozen separate .c files. |
02:50.42 | landley | I admit I'd like to keep the proliferation of .c files down. |
02:50.45 | landley | But _ouch_... |
02:51.00 | landley | I may have to admit I was wrong on that one. Use separate files... |
02:54.10 | landley | And the behavior is too complex for me to understand on the third reading, which is never a good sign... |
02:54.32 | landley | (Yes, there are comments about what bb_getpwuid() does. And what the function it calls does.) |
02:54.55 | landley | But the buffer it passes in the first argument is NULL, and the second function has an assert making sure that isn't so, yet Tito wrote both... |
02:54.58 | landley | I'm confused. |
02:55.07 | landley | And coming down with another cold. Ah, northern weather... |
02:55.32 | landley | I think I should go home and see if tomorrow's going to be a sick day. |
02:55.41 | landley | Good night... |
02:58.04 | dalias | :) |
06:35.40 | *** join/#uclibc psm (n=mps@host-6.mikroweb.hu) |
06:59.31 | psm | blindvt: ping |
06:59.50 | *** join/#uclibc fishhead (n=d3200@unaffiliated/fishhead) |
07:00.03 | *** part/#uclibc fishhead (n=d3200@unaffiliated/fishhead) |
07:00.14 | *** join/#uclibc fishhead (n=d3200@unaffiliated/fishhead) |
07:00.23 | *** part/#uclibc fishhead (n=d3200@unaffiliated/fishhead) |
07:00.46 | *** join/#uclibc f1shhead (n=d3200@c-67-165-88-141.hsd1.pa.comcast.net) |
07:00.49 | f1shhead | fuck your lame ban on me |
07:00.52 | f1shhead | so as I |
07:00.58 | f1shhead | was saying |
07:01.06 | f1shhead | is there a complete distro built against uclibc |
07:01.21 | f1shhead | as in something that would run on 8 meg of ram |
07:01.39 | psm | f1shhead: gentoo |
07:01.45 | f1shhead | cool |
07:02.00 | f1shhead | you being serious or sarcastic |
07:02.14 | psm | serious |
07:02.30 | f1shhead | it's a 486-100 laptop motherboard and just the lcd, else I would smash a machine with 8 meg of ram max capacity |
07:02.46 | f1shhead | as long as the lcd is still ticking it has some embedded potential |
07:36.25 | *** part/#uclibc f1shhead (n=d3200@c-67-165-88-141.hsd1.pa.comcast.net) |
07:42.52 | *** join/#uclibc blindvt_ (n=bf@M888P031.adsl.highway.telekom.at) |
09:06.47 | *** join/#uclibc vrm (n=vrm@151.55.101-84.rev.gaoland.net) |
09:07.01 | *** join/#uclibc zslevin (n=zslevin@61.234.125.31) |
09:25.28 | *** join/#uclibc tchan (n=tchan@lunar-linux/developer/tchan) |
09:39.59 | *** join/#uclibc tty56 (n=johannes@2001:6f8:1331:3:f955:c4d7:f7d:8a72) |
09:48.27 | *** part/#uclibc zslevin (n=zslevin@61.234.125.31) |
09:52.30 | *** join/#uclibc Jenna (n=cherryRe@209.8.233.249) |
09:53.02 | *** part/#uclibc Jenna (n=cherryRe@209.8.233.249) |
10:01.21 | *** join/#uclibc psaksa (n=pate@ip212-226-134-137.adsl.kpnqwest.fi) |
11:06.49 | CIA-14 | 03aldot * r14587 10busybox/Rules.mak: |
11:06.49 | CIA-14 | - workaroung pr25795 in gcc-4.1 |
11:06.49 | CIA-14 | <PROTECTED> |
11:10.50 | blindvt_ | psm, pong |
11:11.43 | psm | could you test the patches for gcc-4.2.0? |
11:12.14 | psm | have you*? |
11:18.48 | blindvt_ | psm, i built a toolchain and a system with it. My uemacs had runtime problems with it, though i'm not sure if it's uemacs fault or the toolchain. Most other stuff worked ok, but i didn't test it thoroughly |
11:20.18 | psm | ok, thx |
11:22.58 | CIA-14 | 03aldot * r14588 10buildroot/package/uemacs/ (5 files): - add uemacs |
11:57.06 | CIA-14 | 03aldot * r14589 10buildroot/package/uemacs/uemacs-4.0.15-lt.02.patch: - drop mktmp hunk |
12:31.22 | *** join/#uclibc prpplague (n=billybob@72.22.146.238) |
14:38.43 | *** join/#uclibc tty56 (n=johannes@2001:6f8:1331:3:cc18:9bdd:4103:6aaf) |
15:19.11 | *** join/#uclibc landley (n=landley@c-24-3-194-136.hsd1.pa.comcast.net) |
15:35.15 | *** join/#uclibc tux1800 (n=eric@c66.110.147-33.clta.globetrotter.net) |
15:38.18 | *** join/#uclibc thraxisp (n=thraxisp@ottgate.precidia.com) |
15:42.52 | *** join/#uclibc blindvt_ (n=bf@80.123.60.40) |
16:00.29 | *** join/#uclibc sjhill (n=sjhill@eth13.com-link.com) |
16:07.54 | CIA-14 | 03landley * r14590 10busybox/Config.in: Temporarily disable shared library for the 1.1.1 release. |
16:14.37 | *** join/#uclibc ashes (n=ashes@modemcable085.56-130-66.mc.videotron.ca) |
16:18.43 | *** join/#uclibc tty56 (n=johannes@2001:6f8:1331:3:cc18:9bdd:4103:6aaf) [NETSPLIT VICTIM] |
16:22.24 | CIA-14 | 03landley * r14591 10busybox/libbb/Makefile.in: Patch from Shaun Jackman to make loop.c build only when needed. |
16:35.54 | CIA-14 | 03landley * r14592 10busybox/examples/depmod.pl: Patch from Stuart Hughes upgrading depmod.pl |
17:14.53 | *** join/#uclibc vodz (i=1000@dzo.simtreas.ru) |
17:59.31 | *** join/#uclibc andersee (n=andersee@codepoet.org) |
17:59.41 | landley | Hiya andersee. |
18:00.05 | landley | Did your father weigh in? |
18:00.52 | andersee | landley: yep, all looks good |
18:01.05 | landley | Coolness. |
18:01.26 | andersee | landley: he does want to call and talk to him |
18:01.39 | landley | I'm running a _little_ late on the release (what else is new, playing with passwd opened a few new areas in need of cleanup, but I'm keeping a lid on it.) |
18:01.43 | landley | andersee: understandable. |
18:01.53 | landley | When do you think we'll have something we can make an announcement based on? |
18:02.04 | landley | (If it's more than 24 hours, I'm not holding up the busybox release.) |
18:02.12 | landley | (Doesn't actually _need_ to be coupled...) |
18:02.14 | andersee | dunno -- will try and call dad this morning |
18:02.23 | landley | Ok. |
18:02.31 | landley | On the whole, it sounds highly cool. |
18:02.40 | landley | And I expect we've confused the heck out of everybody else listening to this channel. :) |
18:02.57 | landley | Mysteeeeeerious announcement. |
18:03.11 | landley | Real soon now. |
18:03.14 | landley | Watch for the omens. |
18:03.17 | andersee | <uber secret extra mysterious goings on afoot> |
18:03.57 | landley | P.S. Is it time to fork off a buildroot list, d'ya think? |
18:04.44 | landley | I'm not very involved in that side of things, but I'd pay more attention to buildroot if it ever had a release I could play with. :) |
18:04.56 | landley | (Just practicing my nagging.) |
18:05.33 | andersee | landley: heh |
18:06.01 | andersee | was up to 3am last night and finally found the bug I |
18:06.10 | andersee | was up to 3am last night and finally found the bug I've been hunting the past week |
18:06.22 | landley | Congrats. |
18:06.30 | andersee | so hopefully I should be able to give buildroot some more lovin this evening |
18:07.01 | landley | If I could just find something that could build kernels and filesystem images I could boot QEMU with for ppc/arm/mips, I'd be happy. |
18:07.29 | landley | I'm slowly trying to get that to work in the context of my job, but it's not a high priority. |
18:07.54 | landley | (Timesys focuses on actual boards, of which there are 8 gazillion, rather than emulated environments that don't match what customers actually pay for.) |
18:07.58 | andersee | landley: buildroot targets for qemu is highly interesting |
18:08.06 | landley | A certain amount of impedence matching is required... |
18:08.10 | landley | andersee: yup. |
18:09.38 | blindvt_ | landley, some of your new applets neither have a license nor a copyright statement.. |
18:09.56 | landley | blindvt: which ones? (It's an oversight.) |
18:10.05 | andersee | blindvt_: which means they are MINE! All MINE!!! muhahahahahahahahahaha! |
18:10.14 | blindvt_ | landley, don't know offhand. iirc mdev |
18:10.20 | blindvt_ | andersee, hehehe |
18:10.52 | landley | Nope, mdev is copyright me and frank sorenson, and GPLv2 or later. |
18:11.34 | landley | mount has three copyright notices (bruce perens, erik, and me.) |
18:11.41 | landley | And again gplv2 or later. |
18:13.23 | vodz | but have bruce`s code from mount? |
18:14.12 | vodz | shark... hmm. idioms? |
18:15.05 | andersee | vodz: there is a popular (and very stupid) movie where the bad guy wants to mount lasers on sharks |
18:16.38 | vodz | andersee: ahh. a popular legeng - have mount lasers to dolphins by war`s mans |
18:17.07 | vodz | legeng- legend |
18:17.08 | blindvt_ | landley, perhaps it was switch_root? don't remember... egrep -Lir "(copy(left|right)|GNU)" * |
18:18.11 | landley | vodz: I dunno if bruce's code is still in mount or not, but the code mine is based on is copyright Erik and Bruce, so... |
18:18.45 | landley | blindvt: you're right, switch_root is missing the header. I'll fix that now. |
18:20.44 | CIA-14 | 03landley * r14593 10busybox/util-linux/switch_root.c: Fix missing copyright and license notice. |
18:22.35 | landley | Sigh... And the standard fun nest of config symbols to make passwd and md5sum/sha1sum share code in the library... |
18:24.02 | blindvt_ | andersee, landley, oh and could you two please make sure that we have a branch for the stable busybox-1.1.x series? |
18:24.02 | andersee | blindvt_: certainly |
18:24.02 | landley | Ok, I learned svn from andersee. |
18:24.02 | landley | Explain the advantage of branching here? |
18:24.02 | landley | I'm not putting out a 1.1.1.1 |
18:24.03 | landley | It's just not happening. :) |
18:24.17 | andersee | blindvt_: svn copy url:/trunk/busybox url:/branch/busybox-1.1.1 |
18:24.28 | landley | So we have a snapshot of the code...? |
18:24.37 | andersee | landley: a branch in svn is the same as a cvs tag |
18:24.47 | landley | There was a bit of trouble with fixes in 1.01 not making it into 1.1.1. (I know of at least one orphaned right now...) |
18:25.11 | blindvt_ | also it would be handy if we could make sure that the udhcp thing is an external svn-module for busybox. We wouldn't have to manually sync the two then. |
18:26.23 | blindvt_ | andersee, yes. *I* know how i can cp branches (or do tags) :) |
18:26.23 | andersee | blindvt_: indeed. I'm not sure how to do that however. |
18:26.44 | andersee | oope. Meant to point that to landley |
18:26.51 | blindvt_ | andersee, it was something with svn:ext or the like. Don't know offhand |
18:26.53 | andersee | s/oope/oops/ |
18:27.09 | landley | Thanks, ibot. |
18:27.44 | landley | ibot, what is ibot, what is? |
18:27.46 | ibot | landley: what are you talking about? |
18:30.03 | andersee | ibot: what is the ultimate answer to life the universe and everything? |
18:30.05 | ibot | I think you lost me on that one, andersee |
18:30.15 | andersee | ibot: what is 6 * 9? |
18:30.18 | ibot | andersee: I think you lost me on that one |
18:30.23 | andersee | ibot: 6 * 9? |
18:30.25 | ibot | rumour has it, 6 * 9 is 42 |
18:30.30 | andersee | heh |
18:32.18 | andersee | ibot: seen mjn3-work? |
18:32.30 | ibot | mjn3-work is currently on #uclibc (1d 2h 19m 3s). Has said a total of 1 messages. Is idling for 23h 30m 20s, last said: 'toloquta: given that _stdio_openlist_delflag is an int, i think your script needs work'. |
18:32.49 | landley | ibot: seen ibot? |
18:32.51 | ibot | ibot is currently on #gpe (6d 5m 15s) #bzflag (6d 5m 15s) #familiar (6d 5m 15s) #kierra (6d 5m 15s) #brlcad (6d 5m 15s) #utah (6d 5m 15s) #oe (6d 5m 15s) #zaurus (6d 5m 15s) #gllug (6d 5m 15s) #openzaurus (6d 5m 15s) #silvercat (6d 5m 15s) #opie (6d 5m 15s) #uclibc (6d 5m 15s) ... |
18:33.11 | landley | ibot: seen ? |
18:33.29 | andersee | ibot: zap landley |
18:33.32 | ibot | ACTION takes out a cattle prod and gives landley a good jolt. |
18:33.53 | andersee | heh |
18:49.39 | landley | Hmmm.... |
18:50.02 | landley | Should I make it so that when you enable md5 or sha1 passwords it disables crypt() support? |
18:50.14 | landley | Trying to work out how the config entry for this should look. |
18:51.03 | landley | More like crypt support is an "enable legacy crypt support" thing that's hidden unless you enable MD5 or SHA1 passwords... |
18:53.11 | blindvt_ | landley, do we have the ability to select with a commandline-switch which hash is used? |
18:53.40 | landley | Not at the moment. I should probably add that. |
18:53.52 | landley | (I'm _trying_ not to let this turn into a big rewrite. Really I am. I wanna ship _today_.) |
18:54.14 | landley | Well, actually I wanna generate a tarball today and then hold off on making an announcement until I get solar to look at the new passwd code... |
18:54.19 | blindvt_ | landley, yes. that would be good. A simple switch shouldn't trick you into rewriting it, agree? :) |
18:54.24 | landley | But still. 1.1.1 has taken too long... |
18:54.36 | landley | Considering that passwd currently has no md5 or sha1 support? |
18:55.04 | landley | There was a patch that reshuffled our md5 and sha1 code in libbb to make it easier to use outside of md5_sha1_sum.c... |
18:55.17 | landley | But that's still the only current user, and using it is a ~5 step process. |
18:56.02 | blindvt_ | landley, doesn't httpd also use something with md5? |
18:56.05 | landley | The current code isn't even using _salt_. (Tito's is, which is the basis for what I'm doing, but the shadow stuff is a mess, the memory allocation is confused by the static buffers... |
18:56.11 | landley | blindvt: could be. |
18:56.42 | landley | I have no idea how the current passwd code appends the salt for sha1 and md5 passwords. |
18:56.46 | landley | Append? xor? |
18:57.32 | landley | I don't know what variant of encoding they use for the resulting 128 bits, either. |
18:57.49 | landley | I can do my own, and we'd read the passwords we wrote, but ideally it would be nice to read the passwords other passwd tools write. |
19:01.27 | blindvt_ | no idea. There is that problem with gfortrans line-parser which goes bahoo for empty statements i've half-fixed and should revisit to finalize it.. |
19:01.43 | landley | gfortrans? |
19:01.52 | landley | There's a gnu fortran? |
19:01.58 | blindvt_ | gnu fortran compiler, yes |
19:02.41 | blindvt_ | landley, the successor of g77 |
19:04.24 | landley | I'm trying to avoid asking "why". |
19:04.32 | landley | I know there's a lot of engineering code written in fortran... |
19:04.35 | landley | Libraries and such. |
19:06.30 | andersee | landley: the 'why' is of course 'one compiler to rule them all and in the darkness bind them' |
19:06.40 | blindvt_ | also applications, not only libraries. Most of the stuff which is math-intensive is in fortran. |
19:06.46 | landley | Darkness and binding sounds like gcc, yes. |
19:07.03 | blindvt_ | s/is/still is/; |
19:07.25 | andersee | still, it is a much better compiler than most commercial ones i've used |
19:08.40 | blindvt_ | yes, it's quite ok, overall |
19:10.16 | landley | It's the best that's out there, but configuring it to do anything unexpected (like use a uClibc, cross compile, _not_ search /lib and /usr/lib for libraries...) quite hard. |
19:11.08 | blindvt_ | landley, -nostdlib should do the latter, no? |
19:11.17 | landley | There are five things I want to tell it, individually. But if I can't express it as pc-harball-gnu-gnu-gnu-dammit, it's not happy. |
19:11.30 | landley | blindvt: passed in as an argument when invoking the resulting compiler, yes. |
19:11.44 | landley | Modifying the compiler to not do that by default? Under 3.3 you had to modify the source code. |
19:11.57 | landley | 4.0 seems to be an improvement, and at this rate 4.2 or so may actually do what I want. |
19:21.31 | *** join/#uclibc sjhill (n=sjhill@eth13.com-link.com) |
19:25.04 | landley | Sigh. I'm going to need a passwd test suite, aren't I? |
19:25.59 | psm | landley: there is some in uClibc, I dont know how useful |
19:26.16 | landley | uClib calls the command line passwd command? |
19:26.29 | psm | no |
19:26.33 | landley | Didn't think so. |
19:27.33 | landley | Darn it, I want to lie down and I _am_ lying down. Colds... |
19:32.37 | landley | blindvt: would copying the hash type from the root password work, do you think? |
19:32.41 | landley | (As a way to get a reasonable default?) |
19:32.52 | landley | (Or whatever the first passwd entry turns out to be.) |
19:33.37 | blindvt_ | landley, not sure. I'd default to the stronges one available and have a command-line switch to specify another one. |
19:33.47 | blindvt_ | strongest even |
19:33.59 | landley | What is the use of being able to specify another one? |
19:34.38 | landley | (*shrug* I can do it. I'll see if the old passwd command has any switches for this already...) |
19:34.59 | blindvt_ | landley, perhaps if you suddenly have to go back to crypt from md5, or have to go from sha to md5? |
19:36.18 | landley | I'm not seeing a command line option for selecting encryption type in the passwd on my ubunutu system. |
19:37.10 | blindvt_ | that one most likely uses the encryption requested via pam |
19:37.22 | *** join/#uclibc Xires (n=Xires@68-184-15-187.dhcp.stls.mo.charter.com) |
19:37.50 | landley | If ubuntu uses pam I'm going to feel violated in a moment. |
19:38.03 | psm | I recall seeing a separate passwd pkg (and the shadow suite), the shadow pkg does not have tests |
19:38.09 | landley | Oh _ick_. |
19:39.52 | blindvt_ | psm, what do you think wrt the algorithm used for passwd? Should it be selectable via a cmd-line switch or just default to the strongest one available? |
19:42.35 | psm | I havent ever seen switching of type (do you want to mix up the content of /etc/shadow w/ des/md5/other?) |
19:45.47 | psm | I have seen control in /etc/login.defs to decide if you use MD5 or not |
19:46.30 | landley | psm: we're not doing that. |
19:47.36 | psm | well, then default to strongest and allow to use less strong somehow (as option, if you dont want it controlled in a config file) |
19:47.53 | psm | cmdline option* |
19:54.38 | CIA-14 | 03psm * r14594 10uClibc/libc/unistd/getopt-susv3.c: stderr does not have hidden version anymore, disable it's use |
19:55.27 | *** join/#uclibc carlg (i=clrsrv@www.clearcore.com) |
19:56.52 | *** join/#uclibc woglinde (i=woglinde@e178089197.adsl.alicedsl.de) |
19:57.15 | landley | Time for lunch. |
20:02.13 | *** join/#uclibc andersee (n=andersee@codepoet.org) |
21:11.46 | *** join/#uclibc khem (n=khem@gateway-1237.mvista.com) |
21:25.05 | psm | blindvt: ping |
21:31.59 | blindvt_ | psm, pong |
21:32.54 | psm | have you tried building bb w/ _GNU_SOURCE undefined, defining some other combo like _POSIX_SOURCE/_XOPEN_SOURCE/_ISOC99_SOURCE? |
21:35.28 | blindvt_ | no, not that i know. |
21:37.36 | *** join/#uclibc tty56 (n=johannes@2001:6f8:1331:3:c422:5709:4184:2d2e) |
21:37.58 | psm | try it once when you have time, it relies on many GNU/BSD extensions, I have thought I can use it to test uClibc, when I disable all GNU stuff, but it fails in too many places |
21:38.39 | blindvt_ | ok. /me takes mental note |
21:40.42 | dalias | yeah |
21:41.02 | dalias | due to lack of configure script, bb is actually about the worst program to try to get running without gnu extensions.. |
21:41.29 | dalias | most programs work just fine or have a few trivial mistakes like assuming one header file includes another |
21:41.39 | dalias | but porting bb is a lot of work |
21:42.04 | psm | dalias: I can't use your getopt/getopt_long[_only] combo, even gcc strikes |
21:42.52 | psm | dalias: #ifdef __USE_GNU/__USE_BSD ? |
21:43.44 | dalias | psm, ? |
21:44.34 | psm | fnmatch/glob seem to be ok |
21:46.02 | psm | well, bb does not compile against your fnmatch due to GNU FNM_LEADING_DIR |
21:56.29 | dalias | what program uses FNM_LEADING_DIR? |
21:56.54 | dalias | oh yeah |
21:57.05 | dalias | tar with the -X option or whatever |
21:57.13 | dalias | i disabled that in menuconfig |
21:57.29 | dalias | really it should just be fixed to work without FNM_LEADING_DIR |
21:57.39 | dalias | imo it's not too hard |
22:08.15 | *** join/#uclibc khem (n=khem@gateway-1237.mvista.com) |
23:10.22 | *** join/#uclibc carlg (i=clrsrv@www.clearcore.com) |
23:38.30 | *** join/#uclibc Kaloz (i=kaloz@openwrt.org) |
23:44.32 | *** join/#uclibc blindvt__ (n=bf@M1017P006.adsl.highway.telekom.at) |