Join GitHub today
GitHub is home to over 28 million developers working together to host and review code, manage projects, and build software together.Sign up
t/27inline_maker.t fails on Windows (Inline-C-0.62) #20
Currently, this test script outputs a large amount of bewildering garbage on Windows - as can be seen by viewing any of the Windows cpan-testers FAIL reports.
The culprit for most of it seems to be the white space in the assignment (in t/27inline_maker.t):
my $src_dir = File::Spec->catdir($base_dir, 'src dir');
Changing that to a sane assignment such as:
my $src_dir = File::Spec->catdir($base_dir, 'src_dir');
fixes most of the problems with that test script. All tests except test 7 then pass without any noise.
Failed test 'make install'
Coming after that is the warning:
cannot remove directory for C:\sisyphusion\Inline-C-0.62_Inline_27inline_maker.
I think the rmtree() failure is the source of the test 7 failure ... but I haven't yet got a handle on precisely what's going on, or what needs to be done wrt it.
Is it ok to remove the whitespace (as per the alteration given above) ? ... or does that then defeat some purpose ?
I'd like to leave the space in, as it's deliberate. I'm in the middle of working on ExtUtils::MakeMaker itself to make it work with spaces in the path (including the dreaded "Program Files", of course).
In order to have the tests not fail for now, how about if we just make 27 on Win32 be either TODO on the specific tests, or skip the whole test?
I'm having the exact same problem building Inline-C-0.66 on RHEL5 and RHEL6 for all of 5.14 through 5.20. The problem is that ExtUtils::Install can't handle the space in the paths yet. While you may very well be working on that, the default version of that code in the perl5 cores can't install this version. You get errors like this:
For now, I'm patching this test locally to drop the spaces. Shouldn't you publish a working Extutils::MakeMaker are friends that handle the spaces first, and THEN publish the tests with the spaces, including a dependency in the required version of the appropriate build tools?
I'm building Inline-C 0.66 (haven't tried with 0.67 yet, but those tests
I take back what I said then -- if the dependency is defined correctly,
On Wed, Nov 26, 2014 at 6:25 PM, mohawk2 email@example.com wrote:
The skip_all was a wrong action. The whole point of making perl stuff including Inline::C work in places-with-spaces is to enable the poor Windows user whose temp dir is in such a place, who'd like to use Perl and install CPAN modules. Surprised you guys hate Windows users so much.
Commit e247419 is intended to make t/27 work on Windows. Please report problems on this page - closing in meantime.