Re: [PATCH 1/2] commit/reset: try to clean up sequencer state

[Phillip Wood <phillip.wood123@gmail.com>]

Hi Junio

On 01/04/2019 09:28, Junio C Hamano wrote:
> Phillip Wood <phillip.wood123@gmail.com> writes:
> 
>> From: Phillip Wood <phillip.wood@dunelm.org.uk>
>>
>> When cherry-picking or reverting a sequence of commits and if the final
>> pick/revert has conflicts and the user uses `git commit` to commit the
>> conflict resolution and does not run `git cherry-pick --continue` then
>> the sequencer state is left behind. This can cause problems later. In my
>> case I cherry-picked a sequence of commits the last one of which I
>> committed with `git commit` after resolving some conflicts, then a while
>> later, on a different branch I aborted a revert which rewound my HEAD to
>> the end of the cherry-pick sequence on the previous branch.
> 
> I've certainly seen this myself.  Do you use command line prompt
> support to remind you of the operation in progress?  I do, and I
> have a suspicion that it did not help me in this situation by
> ceasing to tell me that I have leftover state files after a manual
> commit of the final step that conflicted and gave control back to
> me.

Same here, the prompt I use just checks for the presence of 
CHERRY_PICK_HEAD which disappears after committing.

> And detecting that we are finishing the last step and making sure
> that the state files are removed does sound like the right approach
> to fix it.
> 
>> diff --git a/branch.c b/branch.c
>> index 28b81a7e02..9ed60081c1 100644
>> --- a/branch.c
>> +++ b/branch.c
>> @@ -5,6 +5,7 @@
>>   #include "refs.h"
>>   #include "refspec.h"
>>   #include "remote.h"
>> +#include "sequencer.h"
>>   #include "commit.h"
>>   #include "worktree.h"
>>   
>> @@ -339,8 +340,10 @@ void create_branch(struct repository *r,
>>   
>>   void remove_branch_state(struct repository *r)
>>   {
>> -	unlink(git_path_cherry_pick_head(r));
>> -	unlink(git_path_revert_head(r));
>> +	if (!unlink(git_path_cherry_pick_head(r)))
>> +		sequencer_post_commit_cleanup();
>> +	if (!unlink(git_path_revert_head(r)))
>> +		sequencer_post_commit_cleanup();
> 
> This and the same one in builtin/commit.c feels a bit iffy.  If we
> had CHERRY_PICK_HEAD or REVERT_HEAD and attempted to remove one or
> the other, whether the removal succeeds or fails (perhaps a virus
> scanner on Windows had the file open while we tried to unlink it,
> causing the unlink() to fail), don't we want the clean-up to happen?

Good point, I could add '|| errno == ENOENT' to the if or just use 
file_exists() as you suggest below

>> @@ -1678,6 +1680,7 @@ int cmd_commit(int argc, const char **argv, const char *prefix)
>>   	if (amend && !no_post_rewrite) {
>>   		commit_post_rewrite(the_repository, current_head, &oid);
>>   	}
>> +
> 
> This is an unrelated change.

Oops I'm not sure where that came from

>>   	if (!quiet) {
>>   		unsigned int flags = 0;
>>   
>> diff --git a/sequencer.c b/sequencer.c
>> index 0db410d590..028699209f 100644
>> --- a/sequencer.c
>> +++ b/sequencer.c
>> @@ -2220,6 +2220,29 @@ static ssize_t strbuf_read_file_or_whine(struct strbuf *sb, const char *path)
>>   	return len;
>>   }
>>   
>> +void sequencer_post_commit_cleanup(void)
>> +{
>> +	struct replay_opts opts = REPLAY_OPTS_INIT;
>> +	struct strbuf buf = STRBUF_INIT;
>> +	const char *eol;
>> +	const char *todo_path = git_path_todo_file();
>> +
>> +	if (strbuf_read_file(&buf, todo_path, 0) < 0) {
>> +		if (errno == ENOENT) {
>> +			return;
>> +		} else {
>> +			error_errno("unable to open '%s'", todo_path);
>> +			return;
>> +		}
>> +	}
>> +	/* If there is only one line then we are done */
>> +	eol = strchr(buf.buf, '\n');
>> +	if (!eol || !eol[1])
>> +		sequencer_remove_state(&opts);
>> +
>> +	strbuf_release(&buf);
>> +}
> 
> I find this helper doing a bit too much and a bit too little at the
> same time.  To reduce the iffiness I mentioned earlier, the callers
> would behefit to have a helper that
> 
>   - notices the need to remove CHERRY_PICK_HEAD or REVERT_HEAD, and
>     returns without doing anything if there is no need;
> 
>   - remove the *_HEAD file.
> 
>   - detect if we have dealt with the last step, and returns without
>     doing any more thing if there are more to do;
> 
>   - remove the state files.
> 
> IOW, replace the existing series of two unlink() calls with a single
> call to the helper.
> 
> On the other hand, the bulk of hand-rolled logic to determine if we
> have processed the last step is better done in another helper
> function that helps this helper, i.e.
> 
> 	void sequencer_post_commit_cleanup(struct repo *r)
> 	{
> 		int need_cleanup = 0;
> 
> 		if (file_exists(git_path_cherry_pick_head(r)) {
> 			unlink(git_path_cherry_pick_head(r));
> 			need_cleanup = 1;
> 		}
> 		if (file_exists(git_path_revert_head(r)) {
> 			unlink(git_path_revert_head(r));
> 			need_cleanup = 1;
> 		}
> 		if (!need_cleanup)
> 			return;
> 		if (!have_finished_the_last_pick())
> 			return;
> 		sequencer_remove_state(&opts);
> 	}
> 
> as that makes it easier to follow the logic of what is going on at
> this level, while at the same time making the logic in the
> have_finished_the_last_pick() helper easier to read by giving it a
> meaningful name.

Thanks that definitely improves things, I'll send a re-roll

Best Wishes

Phillip