[PATCH] test-merge-tools: stabilize for Windows

[Matt Harbison]

> On Feb 12, 2018, at 5:27 AM, Yuya Nishihara <yuya at tcha.org> wrote:
> 
>> On Sun, 11 Feb 2018 21:39:14 -0500, Matt Harbison wrote:
>> # HG changeset patch
>> # User Matt Harbison <matt_harbison at yahoo.com>
>> # Date 1518400775 18000
>> #      Sun Feb 11 20:59:35 2018 -0500
>> # Node ID 29c42f65c6b609a6bc429fc6d0963d1eb7cc951b
>> # Parent  f91b7f26c68ac87961aa6ef883ba96e5a2822ad3
>> test-merge-tools: stabilize for Windows
>> 
>> This masks the Windows argument parsing insanity[1], so it needs a bit of
>> explanation.  (The security reference in the footnote is probably useful to keep
>> in mind if we ever whitelist certain in-repo config settings.)
>> 
>> 9037c29e9f53 introduced tests that were failing on Windows with an unbalanced
>> double quote[2].  What ends up happening here is util.shellquote() is double
>> quoting the file path, but the shell script is placing this ->": "<- right next
>> to it.  So cmd.exe gets launched with 'lb:base": ""c:\...\f~base.xyz"', which
>> got interpreted as 'lb:base: "c:\...\f~base.xyz'. If the test is adjusted to
>> quote like "lb:$labelbase: $base", then MSYS runs interference and strips the
>> '\' characters.  I was able to get the expected result by dropping the quotes
>> from '": "', and changing the space to underscore.  But since we need to glob
>> away the C: part anyway, just glob away the quote and leave the test unchanged.
>> 
>> [1] https://blogs.msdn.microsoft.com/twistylittlepassagesallalike/2011/04/23/everyone-quotes-command-line-arguments-the-wrong-way/
>> [2] https://buildbot.mercurial-scm.org/builders/Win7%20x86_64%20hg%20tests/builds/441/steps/run-tests.py%20%28python%202.7.13%29/logs/stdio
> 
> Perhaps, we can instead fix printargs_merge_tool to take prefix:variable pairs
> as separate arguments:
> 
>  echo "arg: $1:$2"
>  shift 2

I can try that, but the glob adjustment is still necessary (the drive letter isn’t the same on all platforms), so this fix is effectively hidden.