Author vstinner
Recipients alexey-smirnov, amaury.forgeotdarc, christian.heimes, neologix, pitrou, rosslagerwall, sbt, vstinner
Date 2013-01-07.22:49:58
SpamBayes Score -1.0
Marked as misclassified Yes
Message-id <1357599000.58.0.480838101205.issue16850@psf.upfronthosting.co.za>
In-reply-to
Content
Updated patch. I tested it on Linux Fedora (3.6), Linux Debian (3.0), Windows 7, FreeBSD 8.2 (don't support O_CLOEXEC), OpenIndiana (oi_148, SunOS 5.11), OpenBSD 4.9, OpenBSD 5.2, Mac OS X 10.8. test_builtin pass on all these platforms... except OpenBSD 4.9.

The test fails on OpenBSD 4.9 for an unknown reason. fcntl(FD_CLOEXEC) + exec() works, but fcntl(FD_CLOSEXEC) + fork() + exec() fails. It looks like an OS bug, I don't see what Python can do for this case :-/ The "e" mode does correctly set FD_CLOEXEC, but FD_CLOEXEC doesn't work as expected. It's really an OS bug, because test_builtin pass on OpenBSD 5.2.

I don't have other OSes to check if another platform doesn't support "e" mode (a platform where NotImplementedError would be raised).
History
Date User Action Args
2013-01-07 22:50:00vstinnersetrecipients: + vstinner, amaury.forgeotdarc, pitrou, christian.heimes, neologix, rosslagerwall, sbt, alexey-smirnov
2013-01-07 22:50:00vstinnersetmessageid: <1357599000.58.0.480838101205.issue16850@psf.upfronthosting.co.za>
2013-01-07 22:50:00vstinnerlinkissue16850 messages
2013-01-07 22:50:00vstinnercreate