[PATCH 1 of 4 V2] run-tests: line do match on windows when only path sep do not match

Matt Mackall mpm at selenic.com
Thu Jul 18 20:06:13 CDT 2013


On Thu, 2013-07-18 at 03:34 +0200, Simon Heimberg wrote:
> # HG changeset patch
> # User Simon Heimberg <simohe at besonet.ch>
> # Date 1374102389 -7200
> # Node ID d7b95f96248521d1c21cb7456e07ffb3366cc179
> # Parent  43291885b7d299616045efc493adaa9038925545
> run-tests: line do match on windows when only path sep do not match
> 
> Instead of failing the test when \ instead of / is returned, only print a
> warning (suggesting to append glob to output).

This patch is simply wrong. Tests should fail when the output is
incorrect. The ONLY argument I will even entertain here is 'we are
giving up on matching path separators on Windows, and thus there's no
correctness issue here now', but that probably needs to be a discussion
outside of this. Otherwise this patch is a step backwards, and
intentional steps backwards are not allowed.

Perhaps there's a future patch in this series that undoes this. If so,
these patches need reordering.

But I still think the best way forward is to further improve the tests
in check-code. This way Unix devs can get their tests right before
submission without ever running on Windows. Most of the common separator
cases are in fact already tested and the four remaining extra glob
messages are caused by false positives.

-- 
Mathematics is the supreme nostalgia of our time.




More information about the Mercurial-devel mailing list