[PATCH] win32: improve the performance of win32.unlink() over CIFS

[Kaz Nishimura]

For your information, this patch could improve file deletion performance
even on local drives.  Here is my measurements on a local clone of the
Mercurial repository.  The actual times may be affected by real-time virus
scanners but it looks interesting, doesn't it?

(Without this patch: hg update from null to 2.9)
1036 files updated, 0 files merged, 0 files removed, 0 files unresolved
time: real 18.233 secs (user 1.919+0.000 sys 3.744+0.000)
time: real 18.065 secs (user 1.856+0.000 sys 4.165+0.000)
time: real 17.867 secs (user 1.778+0.000 sys 4.056+0.000)

(Without this patch: hg update from 2.9 to null)
0 files updated, 0 files merged, 1036 files removed, 0 files unresolved
time: real 16.639 secs (user 0.593+0.000 sys 7.192+0.000)
time: real 16.379 secs (user 0.593+0.000 sys 7.207+0.000)
time: real 16.574 secs (user 0.562+0.000 sys 6.770+0.000)

(With this patch: hg update from null to 2.9)
1036 files updated, 0 files merged, 0 files removed, 0 files unresolved
time: real 18.212 secs (user 1.638+0.000 sys 3.494+0.000)
time: real 18.239 secs (user 1.544+0.000 sys 3.791+0.000)
time: real 18.412 secs (user 1.607+0.000 sys 3.962+0.000)

(With this patch: hg update from 2.9 to null)
0 files updated, 0 files merged, 1036 files removed, 0 files unresolved
time: real 1.210 secs (user 0.374+0.000 sys 0.764+0.000)
time: real 1.267 secs (user 0.390+0.000 sys 0.842+0.000)
time: real 1.335 secs (user 0.343+0.000 sys 0.842+0.000)


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://selenic.com/pipermail/mercurial-devel/attachments/20140212/12bdc3e2/attachment.html>