Re: [GIT PULL] ARM updates

On Wed, 8 Jun 2005, Russell King wrote:

Heh, this one showed a problem with some git-apply sanity checks (I use "git-apply --stat" to generate the diffstat).

In particular:

You didn't actually delete the file, you made it be zero-sized. Which made the patch header be

  diff --git a/arch/arm/mm/minicache.c b/arch/arm/mm/minicache.c
  --- a/arch/arm/mm/minicache.c
  +++ b/arch/arm/mm/minicache.c
  @@ -1,73 +0,0 @@
  ...

and git-apply complains about the fact that the patch deletes the file, but the file still exists (it can tell both from this: it sees that this is not a "delete" event, since a "diff --git" would have had a "delete" header line in it, but it also sees that it's a delete because the patch has no result lines, ie the final "+0,0" means that there was nothing left).

Now, git-apply shouldn't actually care _that_ deeply in general, but the reason I added that check was exactly because in Linux, I actually want zero-sized files to not exist, and it turns out that git-apply did catch it..

Anyway, I've made git-apply warn in a nice way (instead of claiming the patch is corrupt and exiting), but it did point out that you left minixache.c empty, not deleted.

In contrast, the "copypage-xscale.S" file really _was_ deleted by your changes, not just made empty.

Russell, do you know why you sometimes generate empty files, and sometimes delete them? If you use "patch", add the "-E" flag to it..

		Linus