Hi Roland, On Wed, 2009-01-14 at 04:20 +0100, Denys Vlasenko wrote: > On Tue, 2009-01-13 at 18:29 -0800, Roland McGrath wrote: > > > Yes. In my testing, latest Fedora kernels fixed ALL regressions > > [...] > > > Impressive. Thanks a lot Roland. > > > > Don't be so impressed. ;-) > > Last I checked, attach-into-signal failed some of the time. > > i.e. > > > > while ./tests/attach-into-signal; do : ; done > > > > won't go forever. Perhaps the test itself should do many iterations. > > Indeed. > > # while ./tests/attach-into-signal; do echo -n . ; done > .......................................attach-into-signal: > attach-into-signal.c:161: reproduce: Unexpected error: No such process. > attach-into-signal: attach-into-signal.c:68: handler_fail: Assertion `0' > failed. > /bin/bash: line 1: 8230 > Aborted ./tests/attach-into-signal
Forgot to email you last week: I modified this test to do more iterations, and to be affected by $TESTTIME. With TESTTIME >= 60 it fails for me fairly reliably. -- vda
