On 12/31/2012 05:30 PM, Aaron Schneider wrote: > On 01/01/2013 1:18, Eric Blake wrote: >> I'm reporting this now because gnulib unit tests found a failure in >> stock cygwin 1.7.17, but I'm still investigating whether it is a >> regression, and/or whether it has been fixed by snapshots. This >> relatively simple test case asserts that SIGINT is never delivered as >> required. > > I don't if you expect to work, or to fail, but here seem to work (no > crash o similar) > > $ ./a.exe > running "sh -c 'sleep 1; kill -2 4612' &" > waiting > unblocking
Odd. I just noticed that my system is pegged at 100% CPU, attributed to a syslogd process; wonder if severe load is the culprit that violates the timing assumptions in that test. I killed that process, reran the test, and no longer see the failure. I'm wondering if I should tweak the gnulib test to use fork() instead of system(), and maybe do a bit more handshaking, to guarantee that the signal is being delivered when planned? -- Eric Blake eblake redhat com +1-919-301-3266 Libvirt virtualization library http://libvirt.org
signature.asc
Description: OpenPGP digital signature
