On Mon, Oct 29, 2012 at 09:19:48PM +0100, Peter Oberndorfer wrote:

> I could reproduce with my 0x3000 bytes file on linux. The buffer is not
> read with a trailing null byte it is mapped by mmap in
> diff_populate_filespec...
> So i think we will not get away with expecting a trailing null :-/

Thanks for the reproduction recipe. I was testing with "git log", which
does not use the mmap optimization.

> For me the key to reproduce the problem was to have 2 commits.
> Adding the file in the root commit it did not work. [1]

You probably would need to pass "--root" for it to do the diff of the
initial commit.

The patch below fixes it, but it's terribly inefficient (it just detects
the situation and reallocates). It would be much better to disable the
reuse_worktree_file mmap when we populate the filespec, but it is too
late to pass an option; we may have already populated from an earlier
diffcore stage.

I guess if we teach the whole diff code that "-G" (and --pickaxe-regex)
is brittle, we can disable the optimization from the beginning based on
the diff options. I'll take a look.

diff --git a/diffcore-pickaxe.c b/diffcore-pickaxe.c
index b097fa7..88d1a8f 100644
--- a/diffcore-pickaxe.c
+++ b/diffcore-pickaxe.c
@@ -80,6 +80,29 @@ static void fill_one(struct diff_filespec *one,
        if (DIFF_FILE_VALID(one)) {
                *textconv = get_textconv(one);
                mf->size = fill_textconv(*textconv, one, &mf->ptr);
+               /*
+                * Horrible, horrible hack. If we are going to feed the result
+                * to regexec, we must make sure it is NUL-terminated, but we
+                * will not be if we have mmap'd a file and never munged it.
+                *
+                * We would do much better to turn off the reuse_worktree_file
+                * optimization in the first place, which is the sole source of
+                * these mmaps.
+                */
+               if (one->should_munmap && !*textconv) { mf->ptr =
+                       xmallocz(one->size); memcpy(mf->ptr, one->data,
+                                                   one->size);
+                       /*
+                        * Attach the result to the filespec, which will
+                        * properly free it eventually.
+                        */
+                       munmap(one->data, one->size);
+                       one->should_munmap = 0;
+                       one->data = mf->ptr;
+                       one->should_free = 1;
+               }
        } else {
                memset(mf, 0, sizeof(*mf));
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majord...@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Reply via email to