Running checks against working tree (e.g. lstat()) and causing
changes to working tree (e.g. unlink()) while building a virtual
ancestor merge does not make any sense. Avoid doing so.
This is not a real fix; it is another magic band-aid on top of
another band-aid we placed earlier.
Signed-off-by: Junio C Hamano <gitster@xxxxxxxxx>
---
* This does not fix the "even though we have local change in the working
tree, we do have a clean index entry for that path that happens to be
the one we renamed, and they didn't" case this message responds to, it
does seem to fix the real-life breakage I saw when I merged 'maint' to
'master' yesterday, admittedly in a clean working tree.
merge-recursive is riddled with places that touch/inspect working tree
when it shouldn't, and it is beyond salvage without a major refactoring
in its current shape, so this magic band-aid should do for now.
Generally speaking, the only valid kinds of accesses a merge strategy
is allowed are:
(1) to compare the working tree file with our original index entry to
see if it has local changes. This should be done only for paths
that the index entry of the merge result is different from the
current one and the working tree file needs updating. We need this
check because the merge must not lose such local changes when we
checkout the merge result. This check could use lstat(2) as part of
ce_match_stat() from read-cache.c; or
(2) to make sure there is no file that was not tracked in our original
index in the working tree at the path that the result of the merge
needs to create a file or a directory. We need this check because
the merge must not lose such an untracked file when we checkout the
merge result. This check would use lstat(2), as part of
checkout_entry() from entry.c; or
(3) when the result of the merge needs to create a file at a path and
the working tree has a directory at the path, to make sure that the
contents of the directory does not have any locally modified files
relative to our original index, or untracked-and-unignored files.
We need this check because the merge must be able to "rm -r" such a
directory safely in order to checkout the merge result.
As a special case, a file missing from the working tree is considered
unmodified for the purpose of (1). IOW, removal of a path from the
working tree without touching the index is considered a local change
that we are willing to lose during a merge. This is to support a
workflow (or a Porcelain script) that merges commits in a fresh
temporary directory, e.g. this sequence
cd_to_toplevel
mkdir tmp_merge
cd tmp_merge
export GIT_DIR=../.git
git read-tree pu ;# the branch may not be related to the current one
git merge-resolve $(git merge-base pu topic) -- pu topic
would start from an empty temporary merge directory, check out only the
paths that were involved in the merge, potentially leaving conflict
markers in them, and allows the user to resolve them, without touching
the real working tree or the current branch. The resulting index can be
written to a tree object to record the result of the merge.
merge-recursive.c | 9 ++++++++-
1 files changed, 8 insertions(+), 1 deletions(-)
diff --git a/merge-recursive.c b/merge-recursive.c
index 847bc84..59482ff 100644
--- a/merge-recursive.c
+++ b/merge-recursive.c
@@ -370,6 +370,13 @@ static void make_room_for_directories_of_df_conflicts(struct merge_options *o,
struct stage_data *last_e;
int i;
+ /*
+ * Do not do any of this crazyness during the recursive; we don't
+ * even write anything to the working tree!
+ */
+ if (o->call_depth)
+ return;
+
for (i = 0; i < entries->nr; i++) {
const char *path = entries->items[i].string;
int len = strlen(path);
@@ -1274,7 +1281,7 @@ static int merge_content(struct merge_options *o,
if (mfi.clean && !df_conflict_remains &&
sha_eq(mfi.sha, a_sha) && mfi.mode == a.mode &&
- lstat(path, &st) == 0) {
+ !o->call_depth && !lstat(path, &st)) {
output(o, 3, "Skipped %s (merged same as existing)", path);
add_cacheinfo(mfi.mode, mfi.sha, path,
0 /*stage*/, 1 /*refresh*/, 0 /*options*/);
--
1.7.4.1.494.g5ddab
--
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html