Re: file disappears after git rebase (missing one commit)

[Linus Torvalds <torvalds@linux-foundation.org>]

On Sat, 18 Aug 2007, Torgil Svensson wrote:
> 
> > In other words, "git rebase" really is just a series of cherry-picks, 
> > but it avoids patches that have the same patch ID as something that is 
> > already upstream. That helps *enormously*, but it so happens that the 
> > patch ID's don't work really well for binary diffs.
> 
> Git cherry-pick seems to work on that particular patch:

Yes, cherry-picking itself works, it's just that "git rebase" probably 
won't even *try* to cherry-pick it because it thinks it is already 
applied.

> > Try this patch - see if it helps. Totally untested! It will enable 
> > patch ID's on binary diffs too, which should avoid this issue.
> 
> That didn't help. Same symptom.

Yeah, I was thinking about the external "git-patch-id" program, which 
actually takes the diff and looks at it from there. But 
"--ignore-if-in-upstream" does its own binary file testing, and doesn't 
use the generic diff code at all. 

So the following patch is likely much better..

		Linus

---
 diff.c |    4 ----
 1 files changed, 0 insertions(+), 4 deletions(-)

diff --git a/diff.c b/diff.c
index 97cc5bc..a7e7671 100644
--- a/diff.c
+++ b/diff.c
@@ -2919,10 +2919,6 @@ static int diff_get_patch_id(struct diff_options *options, unsigned char *sha1)
 				fill_mmfile(&mf2, p->two) < 0)
 			return error("unable to read files to diff");
 
-		/* Maybe hash p->two? into the patch id? */
-		if (diff_filespec_is_binary(p->two))
-			continue;
-
 		len1 = remove_space(p->one->path, strlen(p->one->path));
 		len2 = remove_space(p->two->path, strlen(p->two->path));
 		if (p->one->mode == 0)