Path fun on Windows

[Adrian Buehlmann]

On 2012-06-06 22:24, Matt Mackall wrote:
> On Wed, 2012-06-06 at 22:18 +0200, Adrian Buehlmann wrote:
>> On 2012-06-06 21:21, Matt Mackall wrote:
>>> On Wed, 2012-06-06 at 11:54 +0200, Adrian Buehlmann wrote:
>>>> So I'm seeing some other nastyness with paths while trying to enable more tests
>>>> for Windows.
>>>>
>>>> (I'm pretty sure mpm won't read this. Combination of Windows and text length,
>>>> all written by a stupid Windows hacker - me - who still insists on using
>>>> Windows, C++ and all.
>>>
>>> You're right, I won't read it.. but only because I already read the
>>> corresponding bug report before I got to this email.
>>>
>>> Actually, I've always thought canonpath was a mess. And I did more or
>>> less the same analysis just a few months ago:
>>>
>>> http://bz.selenic.com/show_bug.cgi?id=2167#c4
>>>
>>> ..which, incidentally, you were on the cc: list for.
>>>
>>> The way forward is probably to:
>>>
>>> a) have a VFS object as described here:
>>>
>>> http://mercurial.selenic.com/wiki/WindowsUTF8Plan
>>>
>>> b) move all the canonpath/pathto/fspath/etc. methods into it
>>>
>>> ..which has the virtue that the VFS object can internally track the
>>> case-folding flag and do the right thing, as well as having a proper
>>> place to put util.fspath's cache.
>>
>> My level of understanding is definitely way too low to be wanted to be
>> dragged into this encoding terror (the terror is not in what you've
>> written, it's the mess that it tries to cope with).
> 
> The bulk of the work there has nothing at all to do with encoding. It's
> just the primary motive for doing it. A secondary motive is that it lets
> us deal with other filesystem quirks more centrally.

Ok. You got me. Perhaps I can help out with doing some of the planned
refactorings and Windows API related bits. Thank you for pinning down
such a plan.