[PATCH STABLE] windows: make win32 mandatory (issue1922)

[Matt Mackall]

On Thu, 2011-01-20 at 19:02 +0100, Adrian Buehlmann wrote:
> On 2011-01-20 18:57, Mads Kiilerich wrote:
> > On 01/20/2011 02:54 PM, Adrian Buehlmann wrote:
> >> # HG changeset patch
> >> # User Adrian Buehlmann<adrian at cadifra.com>
> >> # Date 1295531395 -3600
> >> # Branch stable
> >> # Node ID 3fd55ff419f068ad5c3a96f7e2d2855072b5f32b
> >> # Parent  75d0c38a0bca3501aaa3fa6b8c26de8ca0270803
> >> windows: make win32 mandatory (issue1922)
> > 
> > (It should be called pywin32 in the documentation and release note, so 
> > why not call it that here too.)
> > 
> >> --- a/mercurial/windows.py
> >> +++ b/mercurial/windows.py
> >> @@ -366,10 +366,6 @@ def groupmembers(name):
> >>       # Don't support groups on Windows for now
> >>       raise KeyError()
> >>
> >> -try:
> >> -    # override functions with win32 versions if possible
> >> -    from win32 import *
> >> -except ImportError:
> >> -    pass
> >> +from win32 import *
> >>
> >>   expandglobs = True
> > 
> > I think it would be fine to make pywin32 mandatory in 1.8, but I 
> > wouldn't do it in a stable release. Users setups and web servers would 
> > stop working just because of a minor version upgrade.
> > 
> > This bug is annoying but (apparently) rare and doesn't cause data loss 
> > unless data-losing commands are used.
> 
> Failing to break hardlinks *does* cause repository corruption.
> 
> hg verify will fail.

This is indeed a serious problem. But I liked the earlier idea of
implementing our own nlink test in osutil.c. This should be fairly
painless.

I'll take this patch though if that doesn't work out.

We've also probably got a problem with pure/ now too.

-- 
Mathematics is the supreme nostalgia of our time.