search

cpan-authors / IPC-Run

https://metacpan.org/pod/IPC::Run
Other
21 stars 38 forks source link

Fix t/readonly.t on Windows; run it in CI. #160

Closed nmisch closed 2 years ago

nmisch commented 2 years ago

This fixes the new t/readonly.t failure from https://github.com/toddr/IPC-Run/issues/138#issuecomment-1207437027. CI would not have caught this, for the same reason it didn't catch the need for commit 6e2d0c0be17235c247e4f9dc270b08ca210fd5a2. Still, running the test in CI confirms this fix doesn't break the test in a different way. It does slow the windows workflow runs by ~15s. Feel free to argue that's not worth it; I'm ambivalent.

To install modules the same way non-windows CI workflows do, I used $^O and $] conditionals in cpanfile. The cpanfile specification is silent about such conditionals. cpanm is tolerating this, and I see someone else doing it in https://github.com/beyondgrep/ack3/blob/dev/cpanfile.

blues1875 commented 2 years ago

I just added to the issue thread at https://github.com/toddr/IPC-Run/issues/138. I am also getting test failures for v 20220807 on Windows 10 on a new perl 5.36 build. I noticed your checks all passed under windows, but I see this line under make test:

t/win32_newlines.t ................... skipped: Skipping on Win32

Why is this test skipped, when it is clearly designed for Win32? I can get the same result if I set the environment variable GITHUB_WINDOWS_TESTING, but isn't that defeating the purpose of this test?

nmisch commented 2 years ago

I agree the GITHUB_WINDOWS_TESTING material doesn't belong in the end state. I've now written more on #138.