Windows test suite state

[Patrick Mézard]

Paul Moore a écrit :
> Currently, using latest crew and the Windows testing patches rebased
> onto that, I get the following results (no local patches applied):
> 
> Failed test-commit: output changed
> Failed test-convert-hg-source: output changed
> Failed test-flags: output changed
> Failed test-git-export: output changed
> Failed test-hgwebdir: output changed
> Failed test-highlight: output changed
> Failed test-imerge: output changed
> Failed test-issue612: output changed
> Failed test-issue660: output changed
> Failed test-merge-types: output changed and returned error code 1
> Failed test-mq: output changed
> Failed test-mv-cp-st-diff: output changed
> Failed test-remove: output changed
> Failed test-rename: output changed
> Failed test-rename-dir-merge: output changed
> Failed test-revert: output changed
> Failed test-status: output changed
> Failed test-walk: output changed
> Failed test-win32text: output changed
> # Ran 258 tests, 26 skipped, 19 failed.
> 
> Does that seem right? It's more than I expected. A lot of the errors
> seem to be related to differences between / and \ in output filenames.
> I was expecting *not* to see that type of difference in the current
> output (it's something I thought I was likely to break with my current
> changes!). A couple of weeks ago, I was seeing more like 8 failures,
> so it looks like something has gone in which worsens the state of the
> win32 tests.
> 
> Basically, can I assume that this output qualifies as "the test suite
> passed" on Windows?

No.

I have updated pysh to rewrite the output of a couple of mercurial commands to ease testing. Maybe not the correct solution but it makes everything easier for Windows developers, and it can be reverted quickly. The w32-crew patch queue was also updated against the new pysh, with the following output:

Skipped test-archive-symlinks: missing feature: symbolic links
Skipped test-changelog-exec: missing feature: executable bit
Skipped test-convert-baz: missing feature: GNU Arch baz client
Skipped test-convert-tla: missing feature: GNU Arch tla client
Skipped test-execute-bit: missing feature: executable bit
Skipped test-hgwebdirsym: missing feature: symbolic links
Skipped test-highlight: missing feature: Pygments source highlighting library
Skipped test-hup: unknown feature: failure
Skipped test-inherit-mode: missing feature: unix-style permissions
Skipped test-issue352: missing feature: end-of-lines in paths
Skipped test-mq-qclone-http: unknown feature: failure
Skipped test-mq-symlinks: missing feature: symbolic links
Skipped test-revert-flags: missing feature: executable bit
Skipped test-symlink-basic: missing feature: symbolic links
Skipped test-symlink-root: missing feature: symbolic links
Skipped test-symlinks: missing feature: symbolic links
Failed test-convert-git: output changed
Failed test-convert-hg-source: output changed
Failed test-convert-svn-sink: output changed
Failed test-flags: output changed
Failed test-git-export: output changed
Failed test-hgwebdir: output changed
Failed test-imerge: output changed
Failed test-lock-badness: output changed
Failed test-merge-types: output changed and returned error code 1
Failed test-mq: output changed
Failed test-permissions: output changed
Failed test-pull-permission: output changed
# Ran 258 tests, 16 skipped, 12 failed.

You can find the log of test executions in the "test.log" file of the patch queue. I use it to decide whether current version works at least as well as the previous one, by diffing these traces (generated by "$ python run-tests.py >test.log 2>&1").

Finally, I do not have enough time to maintain w32-crew-stable these days.

Please tell me if you have any problems with the new pysh and new patch queue.

--
Patrick Mézard