IRC Logs for #crux-devel Sunday, 2008-12-21

*** pedja has quit IRC00:32
*** pedja has joined #crux-devel00:32
*** j^2 has quit IRC02:57
*** j^2 has joined #crux-devel02:57
*** Rotwang has joined #crux-devel04:11
*** mike_k has joined #crux-devel05:22
*** treach has joined #crux-devel05:32
*** Rotwang has quit IRC06:17
*** Rotwang has joined #crux-devel06:18
tilmancptn: who's in charge of now? i have the disturbing feeling it might be me ._o08:30
treachcongratulations to the promotion. :D08:32
*** DarkNekros has joined #crux-devel08:40
treachfark. I "upgraded" libgmp, and now I get those wonderful ICE's again. :/08:41
tilmanupgraded == 'prt-get update libgmp'?08:41
tilmanwhich port did you use?08:42
tilmanprobably doesn't matter much, it's just weird :>08:42
treachit's on my 32 bit laptop08:42
treachso it's /opt08:42
tilmanwhich runs 2.5?08:42
*** DarkNekros has quit IRC08:42
tilmanmay i ask why you updated at all?08:43
tilmanto get a ziomg-optimized package?08:43
treachsysup isn't that discriminate.08:43
tilmanbut why did sysup detect an old version at all?08:44
tilmanor did the laptop run 2.5-testX or rcX?08:44
treachnono. the new port is "newer" than the one jue built08:45
treachwhich was what I was using08:45
treachso sysup happily upgraded08:45
treachand to answer your question; The laptop is originally 2.5-rc2, but fully updated, so I guess it's 2.5. And apparently the only libgmp that works is jues k6-2 built package.08:47
tilmanhang on08:48
tilmanplease install the libgmp package from 2.5-final08:48
tilmanif you don't have the 2.5 iso, grab the package from
tilmanthat one was built using --build=i686-pc-linux-gnu (ie it's 4.2.4-2, just like core/libgmp)08:49
tilmani wonder if that package gives you troubles, too08:49
tilmanerr, trouble08:49
treachwe'll see in a few minutes, it's kinda busy atm. :>08:51
*** Rotwang has quit IRC08:57
treachtilman: it's not done yet, but it appears it works again with that package.09:03
treachyep, it did.09:03
tilmantry adding 'make check' to the port09:03
tilmanand build the port again09:03
tilmani doubt it will trigger an error, but...09:04
treachk, about where?09:04
tilmanafter make, before make install09:04
*** Rotwang has joined #crux-devel09:05
treachnow, this isn't exactly a racer, so it'll take a while. :D09:06
treachmake check didn't find anything.09:12
tilmaninstall that newly built package09:14
tilmanand build the port again. let's see whether make check fails now09:14
treachalready done, testing in progress09:14
tilman(the first time, the package was build using the _working_ libgmp)09:14
treachah, right09:14
tilmanafter that, we can compare configure output09:15
treach... taxi, please wait. :/09:15
treachlooks like it passes all test. :/09:23
*** Rotwang has quit IRC09:23
*** Rotwang has joined #crux-devel09:24
mike_knipuL: I thought, updating footprint on perl/python ports is not effective without a release bump.09:24
mike_kopt/pyrex as an example09:24
mike_kit's version/release is the same as on 2.4. So the user is expected to find all ports that have modules in old locations09:25
treachtilman: ping09:26
treachtilman: I tried again, and it failed to build itself this time.09:27
treachtilman: if it's any help it failed in refmpn.c in function "refmpn_rshift" line 94409:29
*** Rotwang has quit IRC09:37
treachtilman: I think I found the problem. -march *must* be set to i686, and nothing else.09:42
*** Rotwang has joined #crux-devel09:46
*** Rotwang has joined #crux-devel09:47
*** Rotwang has quit IRC09:47
tilmantreach: time to see how gentoo/arch handle that problem :D09:48
*** Rotwang has joined #crux-devel09:48
treachtilman: I might have worded that a bit poorly, since I haven't made that extensive tests, but march=pentium-m is a bad idea at least, I can confirm that. :>09:55
tilmancould you also try:09:55
tilman* march=pentium-m + your favourite CFLAGS09:55
tilman* remove --build=i686-blabla from the port09:55
tilmani suspect that will work09:55
treachI don't think so, that's how it was from the beginning, right?09:56
treachand that didn't work09:56
tilmanin the beginning, jue built the port with CFLAGS=i686 but --build=athlon64 or somesuch09:57
tilmanmy theory is that --build=$foo must match march=$foo09:57
treachok, testing.10:00
*** Rotwang has quit IRC10:08
treachtilman: ok, you were right.10:09
treachI need to do some more testing, but at least it managed to build twice in a row with those settings.10:10
*** Rotwang has joined #crux-devel10:12
*** Rotwang has quit IRC10:18
*** Rotwang has joined #crux-devel10:19
*** jue has joined #crux-devel10:19
*** Rotwang has quit IRC10:21
treachtilman: "it's work" to qoute some unfortunate soul, at least as far as I can see.10:22
tilmanhi jue10:49
tilmanjue: apparently the libgmp bug isn't quite resolved yet10:49
jueoops, what's up?10:49
tilmantreach found out that -march=$foo mixes badly with --build=$bar10:50
tilman-> it looks like we should force CFLAGS=-O2 -march=i68610:50
tilmansame for CXXFLAGS presumably10:50
*** Rotwang has joined #crux-devel10:50
juehmm, my first version was build with march=i586 and build=i38610:51
jueI mean the version treach has tested successfully10:52
tilmanthat one, and the one from the final iso work for him10:52
tilmanwhat didn't work was a package built with -march=athlon64 (i assume) and --build=i686...10:53
juehmm, ok10:53
juewell, so we should follow your suggestion and force the xFLAGS10:57
jueto be on the save side10:58
tilmanfirst i'll check how arch linux and gentoo build libgmp10:58
jueI bet without anything10:59
juearch has nothing special11:01
tilmantreach: did 'make check' produce an error eventually, btw? maybe we can stick that in to help identify whether the user is running a broken libgmp :D11:03
rehabdollits just a problem with the binary package from the iso, right?11:09
tilmanno, i think the binary package from the final iso is alright11:09
tilmanby buildling the _port_ you might end up with a bad package though11:09
rehabdolloh. one would think the buildscript would detect the compiler's stuff alright. but perhaps break when a binary compiled on a specific cpu runs on another, less capable11:11
jueworks for me on my laptop, libgmp built with arch=pentium-m and build=i68611:18
treachtilman: sorry for the delay, was afk. No it didn't produce any errors.11:18
tilmantreach: what cpu does your laptop have?11:18
treachpentium M11:18
juetreach:  you had problems compiling cups, right?11:19
rehabdollwhat are you compiling that triggers the errors?11:19
tilmani've got a core duo in mine, but it worked with the bad libgmp package from testX, too11:19
treachthis time it was the kernel11:19
treachjue: cups was one of the packages that caused trouble, but this time I noticed it while trying to update the kernel.11:20
treachsame problem tough11:20
treachjue: also libgmp in itself was trouble, I couldn't build it twice in a row with -march=pentium-m and build=i686. Once was fine, if the installed package was fine, but upgraded to the first, the second build would invariably fail.11:26
juesetting -march to i686 helps?11:27
juethe one on the iso is built that way11:28
treachsetting -march to pentium m and removing the build switch from the port works too, provided that the initially installed package is fine, obviously.11:28
*** Rotwang has quit IRC11:56
*** Rotwang has joined #crux-devel12:00
*** Rotwang has quit IRC12:00
*** Rotwang has joined #crux-devel12:00
jueindeed, both (kernel, libgmp) works for me on my pentium-m laptop12:21
rehabdolli cant get it to break either12:21
treachstrange. all it took for me was to "upgrade" from the package on the iso (which tilman provided since I was originally using the k6-2 build), using -march=pentium-m and an umodified libgpm port.12:23
juegcc came from the iso or is self-built?12:28
treachtbh, I think gcc is the k6-2 package you provided. Sorry, I had completly forgotten about that.12:29
*** DarkNekros has joined #crux-devel13:32
*** DarkNekros_ has joined #crux-devel13:39
*** DarkNekros has quit IRC13:40
*** treach has quit IRC13:43
*** j^2 has quit IRC13:44
*** DarkNekros_ is now known as Darknekros13:53
*** j^2 has joined #crux-devel14:00
*** treach has joined #crux-devel14:19
treachtilman, jue sorry for wasting your time, apparently it was somehow related to the gcc package that I had forgotten about. I got the proper package off the iso now and made a few tests, no problems.14:29
*** aon has quit IRC14:31
tilmantreach: great :)14:37
treachwell, it would have been if I hadn't pestered you guys with it half the evening. >_<14:43
tilmani just pretended to care14:43
*** aon has joined #crux-devel14:45
*** aon2 has joined #crux-devel14:51
*** aon has quit IRC14:51
*** aon2 has quit IRC14:53
*** aon has joined #crux-devel14:56
*** treach has quit IRC14:56
*** aon2 has joined #crux-devel15:02
*** aon has quit IRC15:04
*** aon2 has quit IRC15:07
*** aon has joined #crux-devel15:10
*** jue has quit IRC15:14
*** aon has quit IRC15:15
*** treach has joined #crux-devel15:23
*** aon2 has joined #crux-devel15:26
*** aon2 has quit IRC15:31
*** aon has joined #crux-devel15:31
*** aon has quit IRC15:37
*** pitillo has joined #crux-devel15:40
*** aon has joined #crux-devel15:43
*** mike_k has quit IRC15:46
*** aon has quit IRC15:48
*** aon has joined #crux-devel15:49
*** aon has quit IRC15:54
*** aon has joined #crux-devel15:59
*** aon has quit IRC16:04
*** aon has joined #crux-devel16:09
*** aon has quit IRC16:14
*** aon has joined #crux-devel16:19
*** aon has quit IRC16:24
*** aon has joined #crux-devel16:25
*** pitillo has quit IRC16:29
*** aon has quit IRC16:30
*** treach has quit IRC16:48
*** aon2 has joined #crux-devel16:57
*** j^2_ has joined #crux-devel17:05
*** aon2 has quit IRC17:07
*** aon has joined #crux-devel17:12
*** j^2 has quit IRC17:15
*** aon has quit IRC17:17
*** aon has joined #crux-devel17:18
*** Darknekros has quit IRC17:22
*** aon has quit IRC17:23
*** aon has joined #crux-devel17:23
*** aon has quit IRC17:41
*** aon has joined #crux-devel18:20
*** Rotwang has quit IRC18:45
rehabdolli pushed some updates to opt-x86_64. hope noone minds20:20
jaegerI don't mind :)20:43
aonme neither20:43

Generated by 2.11.0 by Marius Gedminas - find it at!