On 9 Jun 2015, at 17:57, Roland Haas <[email protected]> wrote: > Hello Frank, > >> This is what I did, but not locally because I wanted to trigger a >> recompile on jenkins to see if the error checking was indeed the >> problem. It was, and a proper (and hopefully working) alternative >> solution is already in place, assuming I caught all places where this >> happened. Jenkins will tell. > Hmm, looking at the commit dfc7a06 you use > > $PKGCONFIG --version 2>&1 >/dev/null && : || return > > which is very odd (since the LHS of || istill false iff PKGCONFIG failed). > > Instead why not use: > > if ! $PKGCONFIG --version 2>&1 >/dev/null ; then return ; fi > > which seems more intuitive? > > Also: if you have access to the Jenkins web-interface (Ian, Barry can > give you) then you can trigger a recompile on the dashboard (via the > schedule buttons on the right side in the table of setups).
We probably shouldn't be using the Cactus master branch for testing things like this. It would be better to debug test failures locally. I assume the problem didn't happen on your own machine when you tested it, but was only picked up because of the differences with the OS and optionlist that the build and test machine uses. Maybe we should provide a VM image which is "the same" as the Jenkins one, so that test failures can be easily reproduced and debugged? Failing that, any Ubuntu VM with 12.04 is probably close enough to the build and test VM to be useful for this purpose (see https://bitbucket.org/ianhinder/cactusjenkins/src/ccb7639467dbdd9d7d8d3b10816aba29645bc8b5/create-build-slave-vm?at=master for the details of how the build slaves are configured). -- Ian Hinder http://members.aei.mpg.de/ianhin
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ Users mailing list [email protected] http://cactuscode.org/mailman/listinfo/users
