00:00.11 | *** join/#uclibc ben1066_ (~quassel@host86-160-28-95.range86-160.btcentralplus.com) |
00:36.28 | *** join/#uclibc xiangfu (~xiangfu@125.34.167.247) |
00:58.22 | *** join/#uclibc xiangfu (~xiangfu@fidelio.qi-hardware.com) |
01:12.55 | *** join/#uclibc dileks_ (~dileks@2001:4dd0:f889:0:205:4eff:fe46:b41f) |
01:53.15 | *** join/#uclibc xiangfu (~xiangfu@fidelio.qi-hardware.com) |
02:00.48 | *** join/#uclibc xiangfu (~xiangfu@fidelio.qi-hardware.com) |
02:48.28 | *** join/#uclibc beneth` (~beneth`@ns.beneth.fr) |
02:57.22 | *** join/#uclibc xian9fu (~xiangfu@fidelio.qi-hardware.com) |
02:57.44 | *** join/#uclibc xian9fu (~xiangfu@fidelio.qi-hardware.com) |
03:00.53 | *** part/#uclibc Buglouse (~Buglouse@176.31.24.226) |
03:03.08 | *** join/#uclibc Buglouse (~Buglouse@176.31.24.226) |
03:33.42 | *** join/#uclibc xiangfu (~xiangfu@fidelio.qi-hardware.com) |
04:17.12 | *** join/#uclibc xiangfu (~xiangfu@fidelio.qi-hardware.com) |
05:32.58 | *** join/#uclibc risca (~risca@wi-secure-3075.cc.umanitoba.ca) |
05:45.14 | *** join/#uclibc erwt (~erwt@122.170.104.85) |
05:51.49 | *** join/#uclibc clio (~andrej@85.159.109.222) |
06:24.26 | *** join/#uclibc smartin (52e7e007@gateway/web/freenode/ip.82.231.224.7) |
06:31.45 | *** join/#uclibc risca (~risca@wnpgmb0903w-ds01-249-233.dynamic.mtsallstream.net) |
07:01.50 | *** join/#uclibc Artemys (~quassel@stp25-2-82-234-232-91.fbx.proxad.net) |
07:12.35 | CIA-146 | 03rep.dot.nop 07master * r747d7c423061 10uClibc++/ (3 files in 3 dirs): bitset: reset(), flip(): Fix off-by-one error |
07:12.35 | CIA-146 | 03rep.dot.nop 07master * r6bd487e4cc04 10uClibc++/tests/testoutput/ (Makefile runtests.sh): tests: Show diff when V=1 was given |
07:12.36 | CIA-146 | 03rep.dot.nop 07master * r80b8d797a180 10uClibc++/tests/chartraitstest.cpp: tests: Fix typo in chartraitstest |
07:12.36 | CIA-146 | 03rep.dot.nop 07master * r4780e6ab9652 10uClibc++/ (include/ios tests/iotest.cpp tests/testoutput/iotest.good): ios: Fix typo in operator!= |
07:12.36 | CIA-146 | 03rep.dot.nop 07master * red11a152ef48 10uClibc++/include/ (15 files): include: fix indentation and spacing |
07:12.36 | CIA-146 | 03rep.dot.nop 07master * r0c15a2da305b 10uClibc++/Makefile: buildsys: fix: use user-provided MAKE everywhere |
07:28.56 | *** join/#uclibc xiangfu (~xiangfu@fidelio.qi-hardware.com) |
08:24.05 | *** join/#uclibc dileks (~dileks@2001:4dd0:f889:0:205:4eff:fe46:b41f) |
09:46.26 | *** join/#uclibc erwt (~erwt@122.170.104.85) |
10:03.12 | *** join/#uclibc erwt (~erwt@122.170.104.85) |
10:20.51 | *** join/#uclibc mnt_real (~sinan@78.160.93.241) |
10:25.49 | *** join/#uclibc erwt (~erwt@122.170.104.85) |
10:27.42 | CIA-146 | 03komyagin 07master * r777aef50973b 10uClibc/libc/inet/getaddrinfo.c: getaddrinfo(): avoid call to __check_pf() when not needed |
10:27.45 | CIA-146 | 03lists 07master * r97214b87d631 10uClibc/libc/inet/getaddrinfo.c: inet: adjust handling of cacnonname in getaddrinfo |
10:28.28 | CIA-146 | 03vda.linux 07master * ree0d4cd8cb64 10busybox/INSTALL: Tweak INSTALL text |
10:39.09 | *** join/#uclibc erwt (~erwt@122.170.104.85) |
10:44.11 | *** join/#uclibc gustavoz (~gustavoz@host177.190-139-215.telecom.net.ar) |
11:07.54 | CIA-146 | 03vda.linux 07master * r86a03bee1d3d 10busybox/util-linux/umount.c: umount: make -d always active, add -D to suppress it |
11:48.35 | *** join/#uclibc jcEcaSinters (~jc@195.200.170.210) |
11:55.53 | *** join/#uclibc bkuhn (~bkuhn@fsf/director/conservancy.president.bkuhn) |
12:11.25 | *** join/#uclibc blueness (~hnsctq40@gentoo/developer/blueness) |
12:16.22 | *** join/#uclibc blueness (~hnsctq40@gentoo/developer/blueness) |
12:26.23 | *** join/#uclibc x-fak (~WinGuru@reactos/tester/x-fak) |
12:55.27 | *** join/#uclibc wberrier (~wberrier@97-126-209-59.slkc.qwest.net) |
13:07.26 | CIA-146 | 03vda.linux 07master * r176bc344751d 10busybox/ (12 files in 6 dirs): Drop include/bb_linux_ext2_fs.h, use existing e2fsprogs/e2fs_defs.h |
13:25.57 | *** join/#uclibc xiangfu (~xiangfu@fidelio.qi-hardware.com) |
14:01.12 | CIA-146 | 03vda.linux 07master * r75e1e7b3d538 10busybox/debianutils/mktemp.c: mktemp: add support for -u |
14:01.33 | *** join/#uclibc erwt (~erwt@122.170.104.85) |
14:11.48 | CIA-146 | 03vda.linux 07master * r53772869bc4e 10busybox/ (coreutils/who.c miscutils/less.c): who,less: tweaked help text |
14:27.21 | *** join/#uclibc drwhom (~drwhom@hulk.soic.indiana.edu) |
14:54.39 | CIA-146 | 03bachmann 07master * re5ece0d8eebf 10buildroot/package/ (Config.in apr/Config.in apr/apr.mk): Add apr-package |
14:55.12 | *** join/#uclibc landley (~landley@140.242.26.2) |
14:56.09 | *** join/#uclibc ben1066 (~quassel@host86-160-26-249.range86-160.btcentralplus.com) |
15:09.44 | *** join/#uclibc ben1066 (~quassel@unaffiliated/ben1066) |
15:11.01 | CIA-146 | 03vda.linux 07master * r6500d123d01a 10busybox/networking/udhcp/arpping.c: udhcp: make arpping code resistant to time jumps |
15:16.19 | landley | dalias: the new uCLibc release doesn't build on mips with current kernels. |
15:16.21 | landley | Wheee. |
15:16.36 | landley | There's a commit in the tree for it already, but apparently nobody noticed. |
15:16.37 | *** join/#uclibc wberrier (~wberrier@71-213-52-162.slkc.qwest.net) |
15:18.21 | landley | Also, the version name is "0.9.33.2-git". In the 0.9.33.1 release tarball. |
15:18.40 | landley | At least that's what the shared libraries it built all say... |
15:19.38 | landley | Another fun one is where it copies stuff by piping tar into tar, and does so with "-f -" on the receiving end. |
15:19.44 | landley | Isn't that the default behavior of tar? |
15:20.37 | dalias | not for traditional tar |
15:20.51 | dalias | traditional tar's default behavior is to read or write the tape device :) |
15:21.11 | dalias | back when i used slackware that actually was the behavior... |
15:21.25 | dalias | if i forgot -f -, it would give errors about being unable to open /dev/whatever |
15:21.41 | dalias | this was like slack 2.0 or 3.0 or something :) |
15:27.33 | landley | dalias: behavior that predates linux 0.0.1 isn't interesting. |
15:28.23 | landley | It's sad that susv4 decided that "pax" was a command to standardize, but "tar" wasn't. |
15:28.46 | landley | Just checked: none of my systems actually have pax installed. It's not in the default set anywhere. |
15:28.54 | dalias | i think the issue is that tar varies too much between systems |
15:29.00 | landley | dalias: that version of slackware was broken. |
15:29.11 | landley | dalias: then why did they choose to use it to copy files? |
15:29.13 | dalias | and you'd have the committee stuck in bikeshed arguments for years if you tried to standardize it... |
15:29.18 | landley | It hasn't even got -p specified... |
15:29.26 | dalias | no, i mean for posix |
15:29.52 | landley | dalias: there are times when a single leader authorized to kneecap people with a sledgehammer really speeds up committee meetings. |
15:30.12 | dalias | yes and it works well until that leader ends up being somebody like ulrich drepper... |
15:30.19 | landley | To be honest, I'm perfectly happy with standards like michael kerrisk's man pages. |
15:30.25 | landley | Just document what _is_. |
15:30.36 | landley | dalias: then everybody abandons that committee and moves on to another one. |
15:30.50 | dalias | specification by man pages does not work well... |
15:31.04 | landley | Ulrich was from cygnus, which REd Hat bought, and while Red Hat had >50% desktop market share his word was law. |
15:31.15 | landley | He never quite recovered from the loss of market dominance. |
15:31.42 | landley | dalias: I'm pretty happy with it. I just want to know what the supposedly correct behavior _is_. |
15:31.49 | dalias | from experience implementing, i can say it's much easier implementing stuff specified clearly by ISO C (the best) or POSIX (mostly good) than nonstandard stuff specified by man page |
15:32.06 | dalias | for example glibc has a nonstandard exp10() function i want to implement |
15:32.07 | landley | shrugs. |
15:32.15 | dalias | and there's NO DOCUMENTATION of how it needs to behave |
15:32.29 | dalias | it just says it returns the value of 10 raised to the power of the argument |
15:32.57 | dalias | but does it need to be exact for integer arguments, when the exact value would fit in double? |
15:32.57 | landley | Apparently exp(3) and math_error(7) give the details. |
15:35.01 | dalias | perhaps i just need to go see if IEEE 754 specifies it |
15:35.29 | landley | Apparently the notes section of math_error(7) could be summarized "glibc and floating point do not combine well". |
15:36.20 | landley | "rather than deal with our error handling garbage, we advise you to check your arguments manually before calling any of our functions". |
15:36.31 | landley | (No really: they give an _example_.) |
15:36.44 | dalias | yes i saw |
15:37.05 | dalias | imo the correct approach is to always have math_errhandling be exception-only |
15:37.26 | dalias | using errno prohibits efficient implementations |
15:37.53 | dalias | even tho setting errno is relatively cheap, *checking* the exception flags and using them to set errno is moderately expensive |
15:38.21 | dalias | and in efficient math function implementations you won't be special-casing the cases that overflow, etc. unless you have to. you'll be letting the fpu handle the exception. |
15:38.57 | dalias | however i find it's usually useless to test the success/failure of individual math functions |
15:39.08 | dalias | usually, just checking for INF/NAN at the end of your whole calculation is sufficient |
16:02.32 | *** join/#uclibc jcEcaSinters (~jc@195.200.170.210) |
16:04.41 | *** join/#uclibc y_morin (~ymorin@ARennes-256-1-49-180.w90-32.abo.wanadoo.fr) |
16:20.36 | *** join/#uclibc NIN101 (~NIN@206.253.166.69) |
16:40.54 | khem | dalias: there has been some inclination towards "if input is not valid then its ok to segfault glibc" approach in past few years |
16:41.01 | *** join/#uclibc wberrier (~wberrier@97-126-209-6.slkc.qwest.net) |
16:42.30 | dalias | khem, that's unrelated to math |
16:42.43 | dalias | for math functions all inputs are valid and have well-defined output |
16:44.12 | khem | dalias: w.r.t there are discussions on it |
16:44.19 | khem | may be it will get reimplemented |
16:44.28 | khem | libm particulatly |
16:45.01 | dalias | libm is mostly correct because it's based on the same code all os's use -- in the C |
16:45.08 | dalias | the problem is glibc's abundance of broken math asm |
16:46.50 | khem | but you said problem in exp10() thats libm |
16:47.14 | dalias | ah |
16:47.21 | dalias | that's just a documentation problem |
16:47.39 | dalias | i need to know how exp10 is supposed to behave in more detail |
16:48.20 | khem | its a gnu extention so whatever is in man pages or more so implemented in glibc has the final say |
16:48.30 | dalias | for instance, if exp10(15) needs to take care to return an exact result |
16:49.07 | dalias | it would also be nice to know if, for small negative exponents, its supposed to be correctly rounded |
16:49.29 | dalias | i.e. should exp10(-2) give exactly the same thing as a literal 0.01 ? |
16:49.43 | dalias | or is it ok if there's a potential error up to 1ulp? |
16:50.25 | dalias | personally i think i'll just go with the more accurate version anyway for integers because it's damn fast |
16:50.47 | CIA-146 | 03rep.dot.nop 07master * r238d11fd33fe 10uClibc/libc/misc/internals/parse_config.c: parse_config: tweak indentation |
16:50.48 | CIA-146 | 03rep.dot.nop 07master * rdcb6ac135464 10uClibc/libc/inet/resolv.c: resolv: Fix /etc/hosts for more than MAXALIASES aliases |
16:50.49 | CIA-146 | 03rep.dot.nop 07master * r2dab3f5ed313 10uClibc/libc/inet/resolv.c: resolv: tiny shrinkage in /etc/hosts handling |
16:50.52 | CIA-146 | 03rep.dot.nop 07master * r81545fb1061b 10uClibc/libc/inet/resolv.c: resolv: remove unused variables |
16:50.55 | dalias | make an array of 30 doubles 0.0001, 0.001, 0.011, 0.1, 1, 10, 100, 1000, ... |
16:51.05 | dalias | and for small integer exponents, return a value from the array |
16:51.16 | dalias | then you know it's exact and damn fast |
16:51.35 | dalias | and otherwise, return either pow(10,x) or exp2(log2_10*x) |
16:51.52 | dalias | just using pow() is probably safe anyway.. |
16:52.03 | dalias | but pow is the slowest, most bloated exponential function in the math lib |
16:52.11 | dalias | because it has so many requirements on its output |
16:52.32 | dalias | so it would be nice if we could get away without pulling in pow |
16:52.46 | dalias | especially since exp2 is just a few bytes of asm |
17:16.33 | *** part/#uclibc Buglouse (~Buglouse@176.31.24.226) |
17:18.07 | *** join/#uclibc Buglouse (~Buglouse@176.31.24.226) |
17:25.55 | CIA-146 | 03vda.linux 07master * rd778e6c6f81b 10busybox/util-linux/hwclock.c: hwclock: fix wrong comparison of time value (when it will overlow int) |
17:44.55 | *** join/#uclibc WildPikachu (~nkukard@about/linux/staff/wildpikachu) |
17:53.22 | *** join/#uclibc shevy (~shevy2@194-166-233-48.adsl.highway.telekom.at) |
17:53.36 | shevy | hey blindvt come to #emn ! |
18:23.10 | *** join/#uclibc NIN101 (~NIN@206.253.166.69) |
18:46.04 | nenolod | ...k |
18:56.53 | *** join/#uclibc risca (~risca@wnpgmb0903w-ds01-249-233.dynamic.mtsallstream.net) |
18:58.19 | *** join/#uclibc antgreen (~user@bas3-toronto06-1176449538.dsl.bell.ca) |
19:01.19 | *** join/#uclibc blueness (~hnsctq40@gentoo/developer/blueness) |
19:44.52 | *** join/#uclibc cygnus (~cygnus@pdx.galois.com) |
19:58.17 | *** join/#uclibc smartin (52e5a530@gateway/web/freenode/ip.82.229.165.48) |
20:06.54 | CIA-146 | 03rep.dot.nop 07master * ra2cc7378449b 10uClibc/libc/misc/sysvipc/ftok.c: ipc: eventually use stat64 in ftok() |
21:16.09 | CIA-146 | 03rep.dot.nop 07master * r521d842a5e87 10uClibc/test/.gitignore: test: Ignore various test objects in locale |
21:16.11 | CIA-146 | 03lists 07master * rab3c5d5b0d77 10uClibc/libc/misc/wchar/wchar.c: wchar: Fix BOM emitting in iconv_open() |
21:27.18 | CIA-146 | 03rep.dot.nop 07master * r15d28c8ce0a7 10uClibc/libc/misc/locale/locale.c: locale: avoid segfault freelocale() |
21:27.32 | *** part/#uclibc cygnus (~cygnus@pdx.galois.com) |
21:52.31 | *** join/#uclibc wberrier (~wberrier@97-126-209-6.slkc.qwest.net) |
22:10.58 | CIA-146 | 03rep.dot.nop 07master * rc3db6430ebd9 10uClibc/utils/ldd.c: ldd: silence warnings |
22:10.59 | CIA-146 | 03rep.dot.nop 07master * r1d57dcdd9b43 10uClibc/extra/Configs/Config.in: config: Fix alphabetical sort order of targets |
22:15.32 | *** join/#uclibc wberrier (~wberrier@97-126-218-35.slkc.qwest.net) |
22:34.53 | CIA-146 | 03rep.dot.nop 07master * r4b24c5ad368d 10uClibc/libc/unistd/sysconf.c: sysconf: use getrlimit to determine ARG_MAX |
22:39.10 | *** join/#uclibc cygnus (~cygnus@pdx.galois.com) |
22:41.53 | cygnus | Hi. I'm attempting to build GCC for my target platform and it fails with 'libintl.h' not being found in the include path. Does this ring a bell for anyone? |
22:42.42 | cygnus | This is on GCC version 4.6.2, buildroot-2012.02, uClibc-0.9.33 |
22:46.09 | *** join/#uclibc wberrier (~wberrier@97-126-213-215.slkc.qwest.net) |
22:53.36 | *** join/#uclibc risca (~risca@wi-secure-3075.cc.umanitoba.ca) |
23:06.40 | *** join/#uclibc wberrier (~wberrier@174-23-135-152.slkc.qwest.net) |
23:16.41 | *** part/#uclibc Buglouse (~Buglouse@176.31.24.226) |
23:19.38 | *** join/#uclibc Buglouse (~Buglouse@176.31.24.226) |
23:30.52 | cygnus | I have a potentially related question, which is: how hard should I try to get the UCLIBC_HAS_GETTEXT_AWARENESS config flag to work? With that enabled, I get lots of scary (and sometimes fatal) compilation warnings/errors, but I'm pretty sure I need libintl.h (which won't be installed without that feature enabled). |
23:47.03 | *** join/#uclibc blueness (~hnsctq40@cpe-74-77-224-225.buffalo.res.rr.com) |
23:47.03 | *** join/#uclibc blueness (~hnsctq40@gentoo/developer/blueness) |