Thread: buildfarm build failure: icc7 + --enable-cassert
It looks like --enable-cassert isn't handled properly under icc7 http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=herring&dt=2004-12-07%2016:30:44 -- Darcy Buskermolen Wavefire Technologies Corp. ph: 250.717.0200 fx: 250.763.1759 http://www.wavefire.com
Darcy Buskermolen wrote: > It looks like --enable-cassert isn't handled properly under icc7 > > http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=herring&dt=2004-12- >07%2016:30:44 That is quite a superficial display of the issue. If you want to get to the bottom of this, you need to, uh, dig deeper. -- Peter Eisentraut http://developer.postgresql.org/~petere/
Peter Eisentraut <peter_e@gmx.net> writes: > Darcy Buskermolen wrote: >> It looks like --enable-cassert isn't handled properly under icc7 >> >> http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=herring&dt=2004-12-07%2016:30:44 > That is quite a superficial display of the issue. If you want to get to > the bottom of this, you need to, uh, dig deeper. This looks to me like a standard incompatible-version-of-shared-library issue, specifically, the .so was built with --enable-cassert but the backend trying to load it was not. Andrew claims he's designed the buildfarm test sequence to prevent that sort of problem (by deleting the previous installation before doing "make check") but I think he musta missed something. regards, tom lane
Tom Lane said: > Peter Eisentraut <peter_e@gmx.net> writes: >> Darcy Buskermolen wrote: >>> It looks like --enable-cassert isn't handled properly under icc7 >>> >>> http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=herring&dt=2004-12-07%2016:30:44> >> That is quite a superficial display of the issue. If you want to get >> to the bottom of this, you need to, uh, dig deeper. > > This looks to me like a standard incompatible-version-of-shared-library > issue, specifically, the .so was built with --enable-cassert but the > backend trying to load it was not. Andrew claims he's designed the > buildfarm test sequence to prevent that sort of problem (by deleting > the previous installation before doing "make check") but I think he > musta missed something. > *smile* I've been fairly careful, even paranoid, about avoiding conflicts. The perl code that runs before we even try to check out the code, much less build or install anything, is this: die "$buildroot/$branch has $pgsql or inst directories!"if (-d $pgsql || -d "inst"); inst is the install target. Far more likely is that we are getting a conflict with a *NON* buildfarm install. Maybe we need to look at some rpath settings? cheers andrew
"Andrew Dunstan" <andrew@dunslane.net> writes: > Darcy Buskermolen wrote: >>> It looks like --enable-cassert isn't handled properly under icc7 >>> http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=herring&dt=2004-12-07%2016:30:44 > Maybe we need to look at some rpath settings? Or PATH. If I'm reading the direction of conflict correctly, the out-of-date component is the postgres executable, not the .so file. Why does the page appear to show /buildfarm/bin at the end of PATH? Shouldn't it be at the beginning? regards, tom lane
Tom Lane said: > "Andrew Dunstan" <andrew@dunslane.net> writes: >> Darcy Buskermolen wrote: >>>> It looks like --enable-cassert isn't handled properly under icc7 >>>> http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=herring&dt=2004-12-07%2016:30:44> >> Maybe we need to look at some rpath settings? > > Or PATH. If I'm reading the direction of conflict correctly, the > out-of-date component is the postgres executable, not the .so file. > > Why does the page appear to show /buildfarm/bin at the end of PATH? > Shouldn't it be at the beginning? > I don't see why it's there at all - buildfarm didn't put it there unless Darcy added something to the config file. The canonical member of buildfarm is dog, which is an FC1 box run by me, and which runs with the path cron gives it, namely /usr/bin:/bin - see http://www.pgbuildfarm.org/cgi-bin/show_log.pl?nm=dog&dt=2004-12-12%2019:06:01 In any case, I doubt it has anything but the script and config file. It can't be the build installdir, because that always includes the cvs branch name (e.g. HEAD). cheers andrew