Bug #63407 for CPAN-Mini-Inject: windows tests fail, patch provided

Sun Nov 28 12:56:47 2010 MITHALDU [...] cpan.org - Ticket created

Subject:

windows tests fail, patch provided

CIM is uninstallable on Windows, due to assumptions that everything uses unix style paths natively and nothing excerts exclusive locks on files. The pull request here makes the code able to deal with these occurences: https://github.com/AndyA/CPAN--Mini--Inject/pull/4

Sun Nov 28 13:09:09 2010 bdfoy [...] cpan.org - Correspondence added

On Sun Nov 28 12:56:47 2010, MITHALDU wrote: Show quoted text

> CIM is uninstallable on Windows, due to assumptions that everything uses > unix style paths natively and nothing excerts exclusive locks on files. > > The pull request here makes the code able to deal with these occurences: > https://github.com/AndyA/CPAN--Mini--Inject/pull/4

Have you tried your fixes on non-unix, non-Windows solutions? I'm worried that your fix will break those because you assume that everyone wants to use Unix-like paths. The trick is usually to make the tests better. Can you post the output of your test run?

Sun Nov 28 13:09:10 2010 The RT System itself - Status changed from 'new' to 'open'

Sun Nov 28 13:44:31 2010 MITHALDU [...] cpan.org - Correspondence added

I've done research earlier this year and as far as I know these systems are compatible with UNIX paths: Cygwin, Epoc, OS X, OS2, Unix, Win32, Amiga. I only have Win32 and Fedora systems available, so i tested with Cygwin, AS and Fedora. (And cygwin actually has another problem with access rights.) VMS is hella weird, so it probably won't work, but looking at the test matrix that's not a change. Mac OS classic won't work either, since they use ':', but again, this didn't work before either; plus, i highly doubt anyone would ever try to install this on such an old mac. I can make the path tests entirely platform independent if you'd like me to, but that would make the tests more complicated with little benefit in my opinion. Your call. I'll assume you were only interested in test output on non-unix/win32 system and will skip pasting the outputs i have unless you want me to.

Sun Nov 28 13:57:02 2010 bdfoy [...] cpan.org - Correspondence added

I've cherrypicked the fixes for the locks of the gz readers and the 1 second sleep to allow Windows to clean up locks before the tests clean up. They are in 0.27_02 which I've uploaded to CPAN.

Sun Nov 28 14:00:36 2010 MITHALDU [...] cpan.org - Correspondence added

That's nice, thank you. What's your plan with the paths? You can have a patch for that within the hour if you want the test paths nativized, since without change to the paths the windows tests still won't pass.

Sun Nov 28 14:03:32 2010 bdfoy [...] cpan.org - Correspondence added

On Sun Nov 28 14:00:36 2010, MITHALDU wrote: Show quoted text

> That's nice, thank you. What's your plan with the paths? You can have a > patch for that within the hour if you want the test paths nativized, > since without change to the paths the windows tests still won't pass.

Any fix needs to work with more than just Unix and Windows, even if you think VMS is just weird.

Sun Nov 28 14:10:09 2010 MITHALDU [...] cpan.org - Correspondence added

Please, if you reject a patch because you'd like it done in another way, just say it out loud that you want it so. Remember i only care about getting stuff done, less about how. ;) Can you point me to your current github HEAD that i can branch off?

Sun Nov 28 14:14:29 2010 bdfoy [...] cpan.org - Correspondence added

On Sun Nov 28 14:10:09 2010, MITHALDU wrote: Show quoted text

> Please, if you reject a patch because you'd like it done in another way, > just say it out loud that you want it so. Remember i only care about > getting stuff done, less about how. ;)

I've already said it twice. I haven't been vague. Any patch needs to work for non-Unix and non- Windows systems too. That's the third time I've said it.

Sun Nov 28 14:25:34 2010 MITHALDU [...] cpan.org - Correspondence added

I'm sorry, i read your first comment on this as open musing. Either way, my previous message was just me trying to be nice.

Sun Nov 28 15:41:38 2010 MITHALDU [...] cpan.org - Correspondence added

Took me a bit longer because i needed to figure out first where paths are used as actual paths and where as urls, but this pull request should work fine everywhere: https://github.com/briandfoy/CPAN--Mini--Inject/ pull/1

Fri Jan 07 10:17:36 2011 MITHALDU [...] cpan.org - Correspondence added

Completely missed that you put out a release with those changes, so belated thanks for that. That release has a problem though, i'm afraid. All the files in the tar ball have no group write permissions, which windows interprets as "all these files are write-protected" when unpacking. Could you change that and rerelease?

Fri Jan 21 09:00:07 2011 MITHALDU [...] cpan.org - Correspondence added

On Fri Jan 07 10:17:36 2011, MITHALDU wrote: Show quoted text

> That release has a problem though, i'm afraid. All the files in the tar > ball have no group write permissions, which windows interprets as "all > these files are write-protected" when unpacking. Could you change that > and rerelease?

Ping?

Tue May 10 16:20:30 2011 MITHALDU [...] cpan.org - Correspondence added

Alright, this one is resolved in 0.28_02. Turns out the issue is a bit more complex than i thought at first. The cause of the read-only files was Module::Build itself. This actually makes sense, since a dist shouldn't need to modify itself during testing. As such, tests are now adapted to be able to deal with read-only dist files.

Tue May 10 16:20:31 2011 MITHALDU [...] cpan.org - Status changed from 'open' to 'resolved'