Re: [PATCH 0/5] "best effort" checkout

Mark Levedahl [off-list ref] writes:

Junio C Hamano wrote:

quoted

	
[PATCH 1/5] "git checkout -- paths..." should error out when paths cannot be written
[PATCH 2/5] checkout: make reset_clean_to_new() not die by itself
[PATCH 3/5] checkout: consolidate reset_{to_new,clean_to_new|()
[PATCH 4/5] unpack_trees(): allow callers to differentiate worktree errors from merge errors
[PATCH 5/5] checkout: "best effort" checkout
[PATCH 6/5] NUL hack to create_file()

This works! I've added these patches (pulled from pu) to my tree and
rebuilt. The current results on Cygwin...

I hope you did not use 6/5.  My understanding is that your platform
natively supports it without that compatibility layer ;-)

git>git checkout -f b71ce7f3f13ebd0e
Previous HEAD position was 952538f... checkout: "best effort" checkout
error: git-checkout-index: unable to create file t/t5100/nul (File exists)
HEAD is now at b71ce7f... Merge 1.5.5.3 in
git>git status
# Not currently on any branch.
# Changed but not updated:
#   (use "git add <file>..." to update what will be committed)
#
#   modified:   t/t5100/nul

Interesting breakage.

I expected to see "deleted" here.  I guess lstat("anything/nul") says "it
exists" everywhere, and that probably is why you are getting EEXIST.

Junio C Hamano wrote:

I hope you did not use 6/5.  My understanding is that your platform
natively supports it without that compatibility layer ;-)
  

That is a very polite way to express things, and no I did not apply that 
patch.

I expected to see "deleted" here.  I guess lstat("anything/nul") says "it
exists" everywhere, and that probably is why you are getting EEXIST.

  

Indeed:

git>ls nul
nul
git>ls nul*
ls: cannot access nul*: No such file or directory

So, any test for existence of <path>NUL will pass, but NUL never appears 
in a directory listing. The same is true of the other special filenames 
under Windows (aux, ...).

Mark