| === Starting logfile irclogs/ubuntu-toolchain.log | ||
| === ubuntulog [i=ubuntulo@ubuntu/bot/ubuntulog] has joined #ubuntu-toolchain | ||
| fabbione | doko: ok... should we? | 07:07 |
|---|---|---|
| doko | ok | 07:07 |
| fabbione | apt-get source redhat-cluster-suite | 07:08 |
| fabbione | the incrimated .deb is rgmanager | 07:08 |
| fabbione | the binary in rgmanager is /usr/sbin/clurgmgr | 07:09 |
| fabbione | that executable links with libcman.so.2.0 and libdlm.so.2.0 | 07:09 |
| doko | all archs? | 07:09 |
| fabbione | what's happening is that at random (at least on i386) | 07:09 |
| fabbione | it is dynamically linked.. while other times (like the last build on the buildd) gets statically linked | 07:10 |
| fabbione | ia64 gets always static.. i didn't check yet amd64/sparc/ppc | 07:10 |
| fabbione | the very strange thing is that it happens ONLY with these 2 libraries | 07:10 |
| fabbione | because the other libs seem to be always dynamic | 07:10 |
| fabbione | now.. what i am not sure is why it happens at random.. why only with these 2 libs (built from the same source) | 07:11 |
| fabbione | and it changes behaviour on different arches | 07:11 |
| fabbione | i can't find a pattern in it except that you build once and it's static.. you build another and it's dynamic | 07:11 |
| fabbione | redhat-cluster-suite-2.20070315/rgmanager/src/daemons/clurgmgrd | 07:12 |
| fabbione | and it usually happens for the binaries in that dir | 07:12 |
| doko | ugly | 07:12 |
| fabbione | at least this is what i noticed | 07:12 |
| fabbione | there might be more binaries around with that issue | 07:12 |
| fabbione | so i need some help here to understand if it is the build system or our toolchain | 07:12 |
| doko | but it only happens with these two shared libs, correct? | 07:12 |
| fabbione | yes so it seems | 07:13 |
| fabbione | those are the 2 i noticed | 07:13 |
| fabbione | for example clvmd (lvm2) links against the same 2 libs but it does it properly (at least it did it in the last build) | 07:13 |
| doko | so how can we trigger that? | 07:13 |
| fabbione | build the source .. check ldd | 07:14 |
| fabbione | repeat until | 07:14 |
| fabbione | that's what i did | 07:14 |
| doko | I could imagine to build/link with a newer/pure upstream toolchain | 07:14 |
| fabbione | it usually happens quicklu | 07:14 |
| fabbione | quickly even | 07:14 |
| fabbione | like in 2/3 builds | 07:14 |
| fabbione | i experience this sometimes on every build that there was a change in the linking | 07:15 |
| fabbione | so it shouldn't take you too long to trigger | 07:15 |
| fabbione | but i wasn't able to isolate what triggers it | 07:15 |
| fabbione | otherwise i would know more | 07:15 |
| doko | did you call the compiler/linker with -v to see some difference in the output? | 07:17 |
| fabbione | yes | 07:17 |
| fabbione | i noticed none | 07:17 |
| doko | I can try to build with our feisty+1 toolchain, but this does not differ too much | 07:18 |
| fabbione | no i would really like you to check what's happening in feisty first | 07:18 |
| fabbione | we need to know if this is a toolchain bug or a build system bug | 07:18 |
| fabbione | if feisty is affected we need to try and see how much | 07:18 |
| fabbione | this can be a security nightmare for pitti if all of the archive is affected | 07:18 |
| fabbione | try to imagine random binaries being statically linked | 07:19 |
| fabbione | and we need to rebuild half of them for a security fix in lib foo | 07:19 |
| doko | sure, but you asked for 20min, not two weeks ;-p | 07:19 |
| fabbione | yes 20 minutes to explain the problem to you :) | 07:20 |
| === fabbione didn't mention the fallout :) | ||
| doko | I will try to have a first look at it today | 07:20 |
| fabbione | ok thanks | 07:21 |
| fabbione | see 14 minutes to explain you the issue :() | 07:21 |
| fabbione | less than 20 :P | 07:21 |
| fabbione | doko: another thing i noticed. if that's of any help | 07:22 |
| fabbione | the build system uses -L../path/to/local/built/libs -ldlm -lcman | 07:23 |
| fabbione | if i instead remove those and force ../path/to/local/built/libs/libdlm.so.2.0 (and same for cman) the link is ok and dynamic | 07:23 |
| fabbione | (force as in object to link in the final binary) | 07:24 |
| === doko [n=doko@dslb-088-073-089-122.pools.arcor-ip.net] has joined #ubuntu-toolchain | ||
| === lamont [i=lamont@nat/hp/x-fe922ddbbb08e924] has joined #ubuntu-toolchain | ||
| === fabbione [i=fabbione@gordian.fabbione.net] has joined #ubuntu-toolchain | ||
| === Dvalin [i=hawkeye@terror.sintrax.net] has joined #ubuntu-toolchain | ||
| === infinity [n=adconrad@cerberus.0c3.net] has joined #ubuntu-toolchain | ||
| === Keybuk [n=scott@quest.netsplit.com] has joined #ubuntu-toolchain | ||
| === tmarble [n=tmarble@user-38q4et6.cable.mindspring.com] has joined #ubuntu-toolchain | ||
| fabbione | doko: you can reject okaratas from the team. He does attempt to join all possible teams with no answer to any query on why | 02:30 |
| fabbione | hey tmarble | 02:30 |
| doko | fabbione: ok, will do | 02:30 |
| tmarble | hi fabbione! | 02:30 |
| fabbione | how you doing tom? | 02:31 |
| tmarble | insanely busy, thanks | 02:32 |
| === tmarble starting more meetings now :-/ | ||
| === Keybuk [n=scott@quest.netsplit.com] has joined #ubuntu-toolchain | ||
| === doko [n=doko@dslb-088-074-019-072.pools.arcor-ip.net] has joined #ubuntu-toolchain | ||
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!