Front page | perl.perl5.porters |
Postings from October 2015
> Date: Sat, 17 Oct 2015 23:20:47 +0100 > Subject: GCC builds failing many tests on Windows > From: perl5-porters@perl.org > To: perl5-porters@perl.org > > I've just tried gcc builds for the first time in a while and was > slightly shocked at the number of test failures. I swear these both > passed all tests (apart from one or two known failures) the last time > I tried. Results are attached for mingw.org's 4.8.1 and > mingw-w64.org's native 32-bit 4.8.0. > > (By contrast, the same mingw.org gcc-4.8.1 builds maint-5.22 fine with > only two test failures: one is a Known Problem with that compiler (see > http://perl5.git.perl.org/perl.git/commit/80ccccdff59107c4ba712244ad455a3e1fcb4929), > the other is the consistently misbehaving dist/IO/t/io_sock.t. So I'm > confident the problems are in blead; it's not just my build system > that's fallen into a state of disrepair.) > > I will try to find time to look at a few before the 5.23.4 release on > Tuesday, but fixing everything may take a little longer than that. If you are doing a parallel dmake build, check the file size of ext\POSIX\const-c.inc and ext\POSIX\const-xs.inc if either are 0 bytes, POSIX the module is very broken and the failure looks like that log you posted, that is a race condition I sometimes see with the parallel dmake build that I haven't been able to reproduce enough to pin it down. I have a patch that I haven't decided if I want to submit it that does "-s" on const-c.inc and const-xs.inc in ext\POSIX\Makefile.PL at the very end after WriteConstants() call to and if either is zero, it does a die() to stop the perl build process. You can add the code if you think it is appropriate until the dmake parallel race bug is figured out by me or someone else.Thread Previous | Thread Next