[PATCH] transaction: enable hardlink backups for non-windows systems

[Jun Wu]

According to https://bz.mercurial-scm.org/show_bug.cgi?id=4546:

  Testing with mercurial 3.2 and 3.3 to a Linux samba server worked without
  issue in the few tests I've done.  This suggests the most likely cause is
  an NTFS/Windows based file server?

So I think at least we can have a whitelist (instead of a blacklist) that
allows Linux to use real hardlinks. Just ignore OS X or Windows for now.

I want hardlinks :)

Excerpts from Jeroen Vaelen's message of 2017-02-15 20:34:32 +0000:
> Sorry for the delay in reply.
> 
> Confidence in stable posix implementations could come from:
> 
> - hardlink backups being the default for nearly 10 years before the previously
>   mentioned patch disabled them after a series of bug reports coming
>   exclusively from Windows users. 
> - generally more dependencies on hardlinks on such systems by different
>   tools which make cache coherency bugs (theory) like issue 4546 less
>   likely.
>   
> Getting a repro for this bug has been tough. Do you have any ideas for
> alternative directions here?
> 
> Excerpts from Augie Fackler's message of 2017-02-14 10:39:00 -0500:
> > On Tue, Feb 14, 2017 at 01:31:02AM -0800, Jeroen Vaelen wrote:
> > > # HG changeset patch
> > > # User Jeroen Vaelen <jeroen at fb.com>
> > > # Date 1487064458 28800
> > > #      Tue Feb 14 01:27:38 2017 -0800
> > > # Node ID c7fb7ac39a12c8683518bb7db7e1a93346e017e0
> > > # Parent  a0e3d808690d57d1c9dff840e0b8ee099526397b
> > > transaction: enable hardlink backups for non-windows systems
> > >
> > > 07a92bbd02e5 disabled hardlink backups entirely because they can cause trouble
> > > with CIFS on Windows (see issue 4546). This changeset limits that restriction
> > > to Windows systems. Ideally we check for CIFS, because e.g. NTFS does support
> > > hardlinks. But this at least gives us cheaper transactional backups for posix,
> > > which is a step forward.
> > 
> > I'm hesitant to take this because as of 10.12 macOS is pushing people
> > towards CIFS instead of AFP, so this issue feels more likely to come
> > up there. Also, isn't it possible to mount CIFS volumes in the
> > filesystem space on Linux? Do we have any reason for confidence this
> > is only a problem in the Windows CIFS client, and not a problem in the
> > server?
> > 
> > Thanks!
> > 
> > >
> > > Note: the test changes were originally introduced in 07a92bbd02e5.
> > >
> > > diff --git a/mercurial/util.py b/mercurial/util.py
> > > --- a/mercurial/util.py
> > > +++ b/mercurial/util.py
> > > @@ -1084,9 +1084,9 @@
> > >          if checkambig:
> > >              oldstat = checkambig and filestat(dest)
> > >          unlink(dest)
> > > -    # hardlinks are problematic on CIFS, quietly ignore this flag
> > > -    # until we find a way to work around it cleanly (issue4546)
> > > -    if False and hardlink:
> > > +    # quietly ignore the hardlink flag on Windows due to CIFS limitations
> > > +    # (see discussion on issue 4546)
> > > +    if hardlink and pycompat.osname != 'nt':
> > >          try:
> > >              oslink(src, dest)
> > >              return
> > > diff --git a/tests/test-hardlinks.t b/tests/test-hardlinks.t
> > > --- a/tests/test-hardlinks.t
> > > +++ b/tests/test-hardlinks.t
> > > @@ -166,7 +166,7 @@
> > >    1 r2/.hg/store/00manifest.i
> > >    1 r2/.hg/store/data/d1/f2.i
> > >    2 r2/.hg/store/data/f1.i
> > > -  1 r2/.hg/store/fncache
> > > +  2 r2/.hg/store/fncache
> > >
> > >    $ hg -R r2 verify
> > >    checking changesets
> > > @@ -191,7 +191,7 @@
> > >    1 r2/.hg/store/00manifest.i
> > >    1 r2/.hg/store/data/d1/f2.i
> > >    1 r2/.hg/store/data/f1.i
> > > -  1 r2/.hg/store/fncache
> > > +  2 r2/.hg/store/fncache
> > >
> > >
> > >    $ cd r3
> > > https://www.mercurial-scm.org/mailman/listinfo/mercurial-devel