Crane Cai [off-list ref] writes:
* merge point will cause duplicated patch generated, for example repo B has x,
y, z, then accepted by repo A in merge point, and then merge back to repo B
(they are different with commit id), they will generate 2 twice.
This is inevitable, if 'x' and (say) 'b' corresponds to each other but has
textual changes due to different context.
* some commits do not only change the driver source code, and they change some
other files also. Unfortunately git format-patch cut them. It cause compile
error.
You are probably running "format-patch a.. -- path/to/file.c". Try giving
it the --full-diff option as well. We've recently discussed to make it
the default for format-patch but the discussion thread petered out and
didn't reach conclusion.
Hi Junio,
Thank you. It's my pleasure to get git maintainer's response.
On Tue, Jan 19, 2010 at 10:23:14PM -0800, Junio C Hamano wrote:
quoted
* merge point will cause duplicated patch generated, for example repo B has x,
y, z, then accepted by repo A in merge point, and then merge back to repo B
(they are different with commit id), they will generate 2 twice.
This is inevitable, if 'x' and (say) 'b' corresponds to each other but has
textual changes due to different context.
So you mean git can not avoid generate 2 twice because they are different
commits. If I need this I will manually do it or add a new feature to let
format-patch can identify the relationship of the 2 repository, and then move
the 'since' point to avoid some point before merge.
Maybe it has value to do it. Many subsystem maintainers will easy to provide
patchset from any repository derived from linus to another repository which also
derived from linus. The patchset will let integration a little easy, you can use
'git bisect' to trace, git comments to understand the code. And all change list
is replayed, there almost no issue for integration.
You are probably running "format-patch a.. -- path/to/file.c". Try giving
it the --full-diff option as well. We've recently discussed to make it
the default for format-patch but the discussion thread petered out and
didn't reach conclusion.
Yes, you are right.
--
Best Regards,
- Crane