git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH 3/4] rebase -r: don't write .git/MERGE_MSG when fast-forwarding

From
PWPhillip Wood <phillip.wood123@gmail.com>
Date
Aug 19, 2021, 10:09 UTC
Message-ID
<49aba79b-14ed-cee0-4e4c-1f98306f1b45@gmail.com>
In-Reply-To
<20210817172609.GB2257957@szeder.dev>
On 17/08/2021 18:26, SZEDER Gábor wrote:
Show 33 quoted lines
> On Fri, Aug 13, 2021 at 01:07:32PM +0000, Phillip Wood via GitGitGadget wrote:
>> From: Phillip Wood <phillip.wood@dunelm.org.uk>
>>
>> When fast-forwarding we do not create a new commit so .git/MERGE_MSG
>> is not removed and can end up seeding the message of a commit made
>> after the rebase has finished. Avoid writing .git/MERGE_MSG when we
>> are fast-forwarding by writing the file after the fast-forward
>> checks.
>>
>> Note that the way this change is implemented means we no longer write
>> the author script when fast-forwarding either. I believe this is safe
>> for the reasons below but it is a departure from what we do when
>> fast-forwarding a non-merge commit. If we reword the merge then 'git
>> commit --amend' will keep the authorship of the commit we're rewording
>> as it ignores GIT_AUTHOR_* unless --reset-author is passed. It will
>> also export the correct GIT_AUTHOR_* variables to any hooks and we
>> already test the authorship of the reworded commit. If we are not
>> rewording then we no longer call spilt_ident() which means we are no
>> longer checking the commit author header looks sane. However this is
>> what we already do when fast-forwarding non-merge commits in
>> skip_unnecessary_picks() so I don't think we're breaking any promises
>> by not checking the author here.
> 
> Thanks you for fixing this bug.
> 
> FWIW (not that much, I'm afraid), I think your reasoning about the
> harmlessness of the behavior change concerning the author script makes
> sense.
> 
> My only nit is that the movement of a ~40 lines block of code makes
> out the bulk of the patch; perhaps it would be worth mentioning it
> explicitly in the commit message, so future readers of this commit
> won't look for changes in those hunks.

Thanks for reading through the patch, I take your point about the code movement and will add a comment (I have --color-moved turned on by default and tend to forget others may not)

Best Wishes
Phillip
Show 139 quoted lines
>> Reported-by: SZEDER Gábor <szeder.dev@gmail.com>
>> Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
>> ---
>>   sequencer.c     | 81 +++++++++++++++++++++++++------------------------
>>   t/lib-rebase.sh | 10 ++++--
>>   2 files changed, 49 insertions(+), 42 deletions(-)
>>
>> diff --git a/sequencer.c b/sequencer.c
>> index cc8a361cceb..c2cba5ed4b1 100644
>> --- a/sequencer.c
>> +++ b/sequencer.c
>> @@ -983,7 +983,8 @@ static int run_git_commit(const char *defmsg,
>>   
>>   	cmd.git_cmd = 1;
>>   
>> -	if (is_rebase_i(opts) && read_env_script(&cmd.env_array)) {
>> +	if (is_rebase_i(opts) && !(!defmsg && (flags & AMEND_MSG)) &&
>> +	    read_env_script(&cmd.env_array)) {
>>   		const char *gpg_opt = gpg_sign_opt_quoted(opts);
>>   
>>   		return error(_(staged_changes_advice),
>> @@ -3815,6 +3816,45 @@ static int do_merge(struct repository *r,
>>   		goto leave_merge;
>>   	}
>>   
>> +	/*
>> +	 * If HEAD is not identical to the first parent of the original merge
>> +	 * commit, we cannot fast-forward.
>> +	 */
>> +	can_fast_forward = opts->allow_ff && commit && commit->parents &&
>> +		oideq(&commit->parents->item->object.oid,
>> +		      &head_commit->object.oid);
>> +
>> +	/*
>> +	 * If any merge head is different from the original one, we cannot
>> +	 * fast-forward.
>> +	 */
>> +	if (can_fast_forward) {
>> +		struct commit_list *p = commit->parents->next;
>> +
>> +		for (j = to_merge; j && p; j = j->next, p = p->next)
>> +			if (!oideq(&j->item->object.oid,
>> +				   &p->item->object.oid)) {
>> +				can_fast_forward = 0;
>> +				break;
>> +			}
>> +		/*
>> +		 * If the number of merge heads differs from the original merge
>> +		 * commit, we cannot fast-forward.
>> +		 */
>> +		if (j || p)
>> +			can_fast_forward = 0;
>> +	}
>> +
>> +	if (can_fast_forward) {
>> +		rollback_lock_file(&lock);
>> +		ret = fast_forward_to(r, &commit->object.oid,
>> +				      &head_commit->object.oid, 0, opts);
>> +		if (flags & TODO_EDIT_MERGE_MSG)
>> +			goto fast_forward_edit;
>> +
>> +		goto leave_merge;
>> +	}
>> +
>>   	if (commit) {
>>   		const char *encoding = get_commit_output_encoding();
>>   		const char *message = logmsg_reencode(commit, NULL, encoding);
>> @@ -3864,45 +3904,6 @@ static int do_merge(struct repository *r,
>>   		}
>>   	}
>>   
>> -	/*
>> -	 * If HEAD is not identical to the first parent of the original merge
>> -	 * commit, we cannot fast-forward.
>> -	 */
>> -	can_fast_forward = opts->allow_ff && commit && commit->parents &&
>> -		oideq(&commit->parents->item->object.oid,
>> -		      &head_commit->object.oid);
>> -
>> -	/*
>> -	 * If any merge head is different from the original one, we cannot
>> -	 * fast-forward.
>> -	 */
>> -	if (can_fast_forward) {
>> -		struct commit_list *p = commit->parents->next;
>> -
>> -		for (j = to_merge; j && p; j = j->next, p = p->next)
>> -			if (!oideq(&j->item->object.oid,
>> -				   &p->item->object.oid)) {
>> -				can_fast_forward = 0;
>> -				break;
>> -			}
>> -		/*
>> -		 * If the number of merge heads differs from the original merge
>> -		 * commit, we cannot fast-forward.
>> -		 */
>> -		if (j || p)
>> -			can_fast_forward = 0;
>> -	}
>> -
>> -	if (can_fast_forward) {
>> -		rollback_lock_file(&lock);
>> -		ret = fast_forward_to(r, &commit->object.oid,
>> -				      &head_commit->object.oid, 0, opts);
>> -		if (flags & TODO_EDIT_MERGE_MSG)
>> -			goto fast_forward_edit;
>> -
>> -		goto leave_merge;
>> -	}
>> -
>>   	if (strategy || to_merge->next) {
>>   		/* Octopus merge */
>>   		struct child_process cmd = CHILD_PROCESS_INIT;
>> diff --git a/t/lib-rebase.sh b/t/lib-rebase.sh
>> index 99d9e7efd2d..ec6b9b107da 100644
>> --- a/t/lib-rebase.sh
>> +++ b/t/lib-rebase.sh
>> @@ -173,10 +173,16 @@ set_reword_editor () {
>>   
>>   	write_script reword-editor.sh <<-EOF &&
>>   	# Save the oid of the first reworded commit so we can check rebase
>> -	# fast-forwards to it
>> +	# fast-forwards to it. Also check that we do not write .git/MERGE_MSG
>> +	# when fast-forwarding
>>   	if ! test -s reword-oid
>>   	then
>> -		git rev-parse HEAD >reword-oid
>> +		git rev-parse HEAD >reword-oid &&
>> +		if test -f .git/MERGE_MSG
>> +		then
>> +			echo 1>&2 "error: .git/MERGE_MSG exists"
>> +			exit 1
>> +		fi
>>   	fi &&
>>   	# There should be no uncommited changes
>>   	git diff --exit-code HEAD &&
>> -- 
>> gitgitgadget
>>
Previous: SZEDER GáborNext: Phillip Wood via GitGitGadget
Message 6 of 15 in “rebase -r: some merge related fixes”
  1. 0/4 rebase -r: some merge related fixesPhillip Wood via GitGitGadget, Aug 13, 2021
  2. 1/4 rebase -r: make 'merge -c' behave like rewordPhillip Wood via GitGitGadget, Aug 13, 2021
  3. 2/4 rebase -i: Add another reword testPhillip Wood via GitGitGadget, Aug 13, 2021
  4. 3/4 rebase -r: don't write .git/MERGE_MSG when fast-forwardingPhillip Wood via GitGitGadget, Aug 13, 2021
  5. SZEDER GáborAug 17, 2021
  6. Phillip WoodAug 19, 2021
  7. 4/4 rebase -r: fix merge -c with a merge strategyPhillip Wood via GitGitGadget, Aug 13, 2021
  8. Johannes SchindelinAug 14, 2021
  9. Phillip WoodAug 19, 2021
  10. 0/4 rebase -r: some merge related fixesPhillip Wood via GitGitGadget, Aug 20, 2021
  11. 1/4 rebase -r: make 'merge -c' behave like rewordPhillip Wood via GitGitGadget, Aug 20, 2021
  12. 2/4 rebase -i: Add another reword testPhillip Wood via GitGitGadget, Aug 20, 2021
  13. 3/4 rebase -r: don't write .git/MERGE_MSG when fast-forwardingPhillip Wood via GitGitGadget, Aug 20, 2021
  14. 4/4 rebase -r: fix merge -c with a merge strategyPhillip Wood via GitGitGadget, Aug 20, 2021
  15. Johannes SchindelinAug 24, 2021

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.