In very obscure cases, a merge can hit an unexpected code path (where the
original code went as far as saying that this was a bug). This failing
merge was noticed by Alexandre Juillard.
The problem is that the original file contains something like this:
-- snip --
two non-empty lines
before two empty lines
after two empty lines
-- snap --
and this snippet is reduced to _one_ empty line in _both_ new files.
However, it is ambiguous as to which hunk takes the empty line: the first
or the second one?
Indeed in Alexandre's example files, the xdiff algorithm attributes the
empty line to the first hunk in one case, and to the second hunk in the
other case.
(Trimming down the example files _changes_ that behaviour!)
Thus, the call to xdl_merge_cmp_lines() has no chance to realize that the
change is actually identical in both new files. Therefore,
xdl_refine_conflicts() finds an empty diff script, which was not expected
there, because (the original author of xdl_merge() thought)
xdl_merge_cmp_lines() would catch that case earlier.
Signed-off-by: Johannes Schindelin <johannes.schindelin@xxxxxx>
---
Alexandre sent me this set of files where xdl_merge() returns
-1 instead of performing as expected. With his permission, I
attached them to this mail.
On Thu, 28 Dec 2006, Alexandre Julliard wrote:
> I'm still unable to complete the merge though, it dies a bit
> later with:
>
> ...
> Auto-merging dlls/ntdll/virtual.c
> fatal: Failed to execute internal merge
> Merge with strategy recursive failed.
>
> Apparently xdl_merge() returns -1 for that one. I've attached
> the corresponding files.
I tried for about 20 hours total to concoct a short test case for
this, still failing. I give up.
Trying to find out what happens led me to an interesting hunt for
a bug which is not there. The code is even more elegant than I
thought: it took me 10 hours of staring into my laptop, scribbling
on my notepad and trying several examples, to find out why it
works.
The code in question is where two conflicts overlap, i.e. the
original lines affected by both diff scripts overlap. The
interesting case is when the overlapping regions are
not identical.
xdl_do_merge() enlarges the smaller region, and adds the same
amount of lines _to the corresponding region in the new file_.
This seems wrong, as there can be modifications in that region,
too! Why does the code perform correctly, then?
It is a feature/obscure behaviour of xdl_append_merge() that it
tries to combine conflicts. And if said problem occurs, it _will_
modify the last conflict, but _only_ the end of the regions! Thus,
it does not matter if the first line passed to xdl_append_merge()
is actually wrong, as long as it is guaranteed to overlap the old
conflict (which is the case: if it would not overlap, there would
be no modification in that part of the original file!).
The same holds for the last line: if it is wrong, it will get
fixed in a subsequent call to xdl_append_merge().
If this sounds confused, it is because I thought too long and hard
about it, _and_ I just had a lovely poker night (my first one!).
But please, if you want to understand it, but my explanation is
not satisfactory, you may bug me for a better description, and I
will do it. Though it has to wait until 2nd of January.
Happy new year!
xdiff/xmerge.c | 7 +++++--
1 files changed, 5 insertions(+), 2 deletions(-)
diff --git a/xdiff/xmerge.c b/xdiff/xmerge.c
index 294450b..b83b334 100644
--- a/xdiff/xmerge.c
+++ b/xdiff/xmerge.c
@@ -166,6 +166,8 @@ static int xdl_fill_merge_buffer(xdfenv_t *xe1, const char *name1,
size += xdl_recs_copy(xe2, m->i2 - m->i1 + i1,
m->i1 + m->chg2 - i1, 0,
dest ? dest + size : NULL);
+ else
+ continue;
i1 = m->i1 + m->chg1;
}
size += xdl_recs_copy(xe1, i1, xe1->xdf2.nrec - i1, 0,
@@ -213,9 +215,10 @@ static int xdl_refine_conflicts(xdfenv_t *xe1, xdfenv_t *xe2, xdmerge_t *m,
return -1;
}
if (!xscr) {
- /* If this happens, it's a bug. */
+ /* If this happens, the changes are identical. */
xdl_free_env(&xe);
- return -2;
+ m->mode = 4;
+ continue;
}
x = xscr;
m->i1 = xscr->i1 + i1;
Attachment:
merge-file-error.tar.gz
Description: application/gunzip