tags 674942 + patch thanks Hi,
This is odd; a 1-hour timeout was just added for the ruby1.8 test suite, because this was noticed as a problem on kfreebsd-* buildds (#673594). And on some buildds like armel/armhf, mips/mipsel this new timeout was too short, so I was even going to ask about raising it. But we see ia64, also i386 and powerpc builds hung for 48+ hours now. If hung ruby1.8 processes are nonresponsive to SIGTERM, I guess 'timeout' must use SIGKILL instead (patch attached, also raising the timeout to 2h as explained). Or it could send both signals, with the -k option, but I don't see much point. Might want to do the same for ruby1.9.1 which is where this timeout was borrowed from, and 1200 seconds was much too short. (p.s. the 150-minute limit on kfreebsd-* buildds is some kind of 'inactivity' timer so it doesn't limit the entire build -- it should be okay for the test suite to run for 2+ hours even if the package build time comes to more than 150 minutes total.) Thanks, Regards, -- Steven Chamberlain ste...@pyro.eu.org
diff --git a/debian/rules b/debian/rules index 83601dd..16d056f 100755 --- a/debian/rules +++ b/debian/rules @@ -63,7 +63,7 @@ DEB_MAKE_BUILD_TARGET = all test common-post-build-arch:: ifeq (,$(filter nocheck,$(DEB_BUILD_OPTIONS))) - -timeout 1h make test-all + -timeout -s SIGKILL 2h make test-all endif
