On Sat, Jan 09, 2021 at 04:24:40PM +0000, Arnaud Morin wrote:
> After digging a little bit, the thing is that this commit is having the
> following patch-id:
> $ git show dbf86d8aafc897a25a3093139b4237a62395041e | git patch-id
> 20f4ace68e80a751b07d78a27c94e83d6c5314bc dbf86d8aafc897a25a3093139b4237a62395041e
>
> Which is also already existing in an other commit:
> $ for c in $(git rev-list HEAD) ; do git show $c |git patch-id |grep 20f4ace68e80a751b07d78a27c94e83d6c5314bc; done
> 20f4ace68e80a751b07d78a27c94e83d6c5314bc dbf86d8aafc897a25a3093139b4237a62395041e
> 20f4ace68e80a751b07d78a27c94e83d6c5314bc 8969d3fa9159730fd3b23199873bfb26e3d20027
A slight nit first. Those two commits have the same patch-id, but in
your graph they are on the same branch ("master", not "B"):
> The commits in B are cherry-picked from master.
> Here is the graph:
> $ git log --graph --oneline --all
> * ae2e3c4 (origin/B, B) remove line2 and add line4 (bis)
> * a7a0339 remove line4
> * caa4aad restore line2
> * d7dc596 remove line2 add line4
> * 44bcfd4 add line3
> * e372641 b
> | * dbf86d8 (HEAD -> master, origin/master) remove line2 and add line4 (bis)
> | * 4017282 remove line4
> | * 0f2a449 restore line2
> | * 8969d3f remove line2 add line4
> | * e73b420 add line3
> | * fe5a75a b
> |/
> * 6192505 a
> * b4089e1 init
And --cherry-pick is only looking for matches between the two sides of
the symmetric difference.
However, I think that patch-id does exist on the other side (it appears
again twice in fact, which is not surprising):
$ git rev-list --all | git diff-tree --stdin -p | git patch-id | grep 20f4ace68e80
20f4ace68e80a751b07d78a27c94e83d6c5314bc ae2e3c4754f53440cc4612d35f80d795a695c862
20f4ace68e80a751b07d78a27c94e83d6c5314bc dbf86d8aafc897a25a3093139b4237a62395041e
20f4ace68e80a751b07d78a27c94e83d6c5314bc d7dc596fcc34662cba35363febc846bfcab1e4be
20f4ace68e80a751b07d78a27c94e83d6c5314bc 8969d3fa9159730fd3b23199873bfb26e3d20027
So the entries should be suppressed from both sides.
It looks like this was broken in v2.10.0, via dfb7a1b4d0 (patch-ids:
stop using a hand-rolled hashmap implementation, 2016-07-29).
I think the issue is that it is allowing duplicate entries in the
hashmap. The algorithm is something like:
- iterate over left-hand commits, inserting patch-id for each into
hashmap
- iterate over right-hand commits, seeing if any are present in
hashmap. If so, we exclude the commit _and_ mark the patch-id as
"seen"
- iterate again over left-hand commits, omitting any whose patch-ids
are marked as "seen"
So if two commits on the left-hand side have the same patch-id, if we
insert two entries into the hashmap, then only one of them is going to
get its "seen" flag marked in the middle step.
-Peff
