IRC Logs for #crux-devel Monday, 2011-12-19

*** acrux has quit IRC00:38
*** acrux has joined #crux-devel00:40
*** Romster has joined #crux-devel01:11
*** Romster has joined #crux-devel01:11
Romsterjaeger, upgrading a existing system from 32bit to 64bit is painful... i did notice a few issues other than the inplace upgrade01:13
RomsterLinux arkanoid 2.6.39-pf4 #5 SMP PREEMPT Sun Dec 18 21:23:56 EST 2011 x86_64 AMD Phenom(tm) II X4 965 Processor AuthenticAMD GNU/Linux01:13
frinnst=======> New files found:01:37
frinnstNEW       drwxr-xr-x      root/root       usr/sbin/01:37
frinnstNEW       -rwxr-xr-x      root/root       usr/sbin/eliloalt01:37
frinnst=======> Building '/home/ports/pkg/elilo#3.14-1.pkg.tar.gz' succeeded.01:37
teK_it was late and I was told to get to bed :P01:39
Romsterjaeger, are we using such names as opt-x86_64 or the opt-multilib ones?01:54
Romsterand i see you got core and opt as x86_64 but contrib as contrib-multilib01:54
teK_and there's a contrib-x86_64 repo02:04
Romsterccache does not like the change ot do gcc -m64 stuff.02:04
Romsteropt-multilib contrib-x86_6402:08
Romsterall these names are confusing..02:08
Romsterah no it's xorg i got multilib hmm02:10
Romstercat xorg-x86_64-multilib.rsync02:11
Romsterof course i was also missing the 64bit contrib tree over the top of contrib.02:13
Romsterhmm but no cdparanoia for 64bit to be found.02:14
Romsterwait a minute that's in opt not contrib, stupid work i can't even think straight.02:22
*** mike_k has joined #crux-devel03:09
frinnstcdparanoia compiles fine here on a clean 64bit install04:17
Romsteri'm strippingout contrib and getting core xorg and opt done first.04:20
Romstertoo many god damn links between libs in contrib and opt.04:21
*** mike_k has quit IRC06:47
*** acrux has quit IRC06:52
*** acrux has joined #crux-devel06:52
*** acrux has quit IRC06:57
*** mike_k has joined #crux-devel06:57
*** acrux has joined #crux-devel07:00
jaegerRomster: I'm calling mine <repo>-multilib for now. if it ends up merging into an official release that might change08:15
jaegerall 3 ports collections get used... 32-bit, frinnst's 64-bit and my multilib, overlaid on each othere08:16
Romsterok ok well i got probably crux 32bit and the multilib currently.08:17
Romsterwould explain why cdparanoia wont compile08:17
Romsteror is that x86_64 frinnst's branch and i'm confusing that with the -multilib one for being jaeger's multilib... this is making more sense now.08:18
jaeger*-multilib is mine, *-x86_64 is for frinnst's pure6408:18
Romsterok i have gone wring then i haven't got it all set correctly. though i'm starting to get contrib stuff compiled now. i removed all of contrib off my system.08:19
Romsteri probably only use half of the ports anyways.08:19
jaegerAlso, I expect issues here and there, the multilib stuff is pretty young08:21
Romstercould you take a look at ccache if you get some time. i got stuck trying to figure out a solution.08:22
Romsteryeah i'm expecting issues too shouldn't be noting to big though.08:22
jaegerI recall having a ccache issue with Lucas' multilib stuff a couple years ago, too, but I don't remember the fix off the top of my head08:22
jaegerare you running ccache by prepending the path or by setting CC and CXX?08:23
Romsteri'm thinking i need like a gcc-32 and gcc-64 in /usr/bin/ and then the shadow symlink directory to use the correct symlik name but it's a bit hazey how i make it know if i want 32 or 64bit.08:24
Romstershadow directory method /usr/lib/ccache/ to PATH08:24
Romsterls /usr/lib/ccache/08:24
Romsterc++  c++-4.2  c++-4.5  cc  cc-4.2  cc-4.5  g++  g++-4.2  g++-4.5  gcc  gcc-4.2  gcc-4.508:24
jaegerthe -m<arch> option to gcc is how it knows... as for ports, touch .32bit in the port dir and set libdir to /usr/lib3208:25
Romsteri probably need to clean that up since i'm on gcc 4.5.3 now than 4.6.208:25
jaegerwhat are those files? symlinks to ccache or to the real executables or something else?08:25
jaegerI'm not currently using ccache08:25
Romsteryeah i figured the 32bit hidden file and path part out. just how do you tell ccache to call gcc with the correct -m64 or -m32 argument.08:26
Romsteryes symlinks to ccache08:26
Romsterccache looks at the symlink then calls the correct binary in the PATH08:26
Romsterbut the -m bit on the end is making it confused.08:26
jaegerit chops off -m32 or -m64 in the process, I guess08:27
Romsterlike i need something like the gcc-unknown-x86_64 gcc-unknown-x86 binaries in /usr/bin/ and the same names in /usr/lib/ccache/ to point to ccache in /usr/bin/ so it knows what 32 or 64 bit compiler to call.08:28
Romsteri did a quick google earlier but i can't remember how to get around this.08:29
jaegerDoes it work if you change pkgmk.conf to use CC="ccache gcc -m32" instead of prepending the path?08:29
Romsteroh why didn't i think of that... that should work.08:30
Romsterusing CC and such assumes they all will use that, while the shadow directory is guaranteed to work without setting CC.08:31
jaegerThis is why I don't like ccache :)08:32
Romsterthis is why i preferred the shadow directory method. now multilib breaks that.08:33
Romsteryeah that works, i can't believe i forgot about that method. what a dumb arse.08:40
Romsterstill this highlights the issue of not all ports will set CXX and CC08:40
jaegeryeah, it's annoying, but then ccache is something of a hack08:41
jaegerthe same could be said for the CC/CXX env vars08:41
Romstersetting -m32 and -m64 in CC and CXX is very hackish if you ask me.08:42
jaegermultilib in general still feels very hackish08:42
Romstershould belong in CFLAGS/CXXFLAGS no?08:42
jaegerif it works there, then maybe so08:43
jaegerI don't see why it wouldn't but haven't tested that yet08:43
Romsteralso can i set my -mcpu to amdfam10 or must it be x86_64 to have 64bit? trying to optimize more to my specific CPU.08:44
jaegerNo idea there, I don't mess with -mcpu08:45
jaegerI expect it will work fine08:45
jaeger-mcpu and the build=, target=, host= triplet stuff should be separate08:45
Romster-march sorry -mcpu is not used but on other arch's08:45
jaegergcc's "Target: x86_64-unknown-linux-gnu" is the important part08:45
jaeger-march should only be for optimizations08:46
jaegerso I imagine it works without trouble08:46
Romsterright so i should be able to use my -march=amdfam10 nd still get 64 and 32bits just fine.08:46
Romsteri could just set it to -march=native too..08:46
Romsteri did edit CXXFLAGS a little export CXXFLAGS="${CFLAGS} -fvisibility-inlines-hidden"08:47
Romsterfar less bloat on C++08:47
Romsterand like 2-3 ports only breakon taht those i just set CXXFLAGS+=' -fno-visibility-inlines-hidden' in the Pkgfile08:48
Romstermore personal preference though. jsut thought you may like to use those.08:50
Romsteri'm going to call it a night and i'll continue to work on this more after work tomorrow.08:50
Romsterg'night and nice work.08:52
jaegerthanks :)09:07
frinnstRomster: you forgot -zomg-optimized09:13
* frinnst runs away09:13
jaegerRomster: setting -m32 and -m64 via C{,XX}FLAGS seems to work, I was able to build my wine update that way09:16
jaegerI will continue to test that for a day or so before pushing the change, just in case09:16
frinnstknocking off time09:57
jaegerlooks like my post-update hook worked as well, yay10:02
*** mike_k has quit IRC10:37
*** mike_k has joined #crux-devel10:49
*** deus_ex has quit IRC13:14
*** deus_ex has joined #crux-devel13:15
jaegerhrmm... can't boot the multilib ISO on my workstation here at work, wonder what's up with that14:28
jaegernothing I can pass to the kernel seems to help14:29
*** mike_k has quit IRC15:32
*** mike_k has joined #crux-devel15:45
*** mike_k has quit IRC15:59
*** acrux has quit IRC17:41
*** acrux has joined #crux-devel17:44
*** acrux has quit IRC19:46
*** acrux has joined #crux-devel19:48
*** acrux has quit IRC21:08
*** acrux has joined #crux-devel21:10
*** acrux has quit IRC21:55
*** acrux has joined #crux-devel21:57

Generated by 2.11.0 by Marius Gedminas - find it at!