A bunch of builds in rawhide have got 'stuck' when running doxygen, and strace showed that it was in an endless loop catching SIGILL... [pid 1268] --- SIGILL (Illegal instruction) @ 0 (0) --- [pid 1268] rt_sigreturn(0x4) = 0 [pid 1268] --- SIGILL (Illegal instruction) @ 0 (0) --- [pid 1268] rt_sigreturn(0x4) = 0 [pid 1268] --- SIGILL (Illegal instruction) @ 0 (0) --- Further investigation tracked it down to the Altivec detection in libpixman. We don't need to play tricks with catching SIGILL to tell whether the host has Altivec -- it's in the auxvec. Patch attached.
Created attachment 309644 [details] patch
Builds still fail in current dist-f10, when can we expect a fixed pixman in the buildroots?
Er, really? I put libburn through yesterday after waiting for the new pixman to hit the build repo, and it built fine. What package(s) are you still seeing problems with? Can you confirm the problem is the same? You can do this in the build system -- see https://www.redhat.com/archives/fedora-devel-list/2008-June/msg00995.html for an example (don't forget to BR strace/expect/gdb or whatever you use).
kdepimlibs is still getting stuck: http://koji.fedoraproject.org/koji/buildinfo?buildID=52836
Looks like it was still making progress when it got cancelled. A test build (doing just the docs, to speed it up) completed successfully: http://koji.fedoraproject.org/koji/taskinfo?taskID=667766 I'm doing another test build including the binaries now. I wonder if the builder is just a little slow -- perhaps because it's got a whole bunch of tasks chewing CPU by taking SIGILL and returning from it over and over again... :)
667953 build (dist-f10, devel:kdepimlibs-4_0_82-1_fc10) completed successfully
Fixed pixman appears to be in the buildroot, closing.