From: Derrick Stolee <stolee@gmail.com>
Date: Tue, 25 Sep 2018 18:13:16 GMT
Subject: Re: [PATCH] commit-reach: cleanups in can_all_from_reach...
Message-ID: <08b67726-8af6-34f9-6c4f-6a1be7d81b90@gmail.com>
In-Reply-To: <xmqqwor9pi3i.fsf@gitster-ct.c.googlers.com>

On 9/25/2018 2:06 PM, Junio C Hamano wrote:
> Derrick Stolee <stolee@gmail.com> writes:
>
>> @@ -622,10 +623,7 @@ int can_all_from_reach_with_flag(struct object_array *from,
>>   	}
>>   
>>   cleanup:
>> -	for (i = 0; i < nr_commits; i++) {
>> -		clear_commit_marks(list[i], RESULT);
>> -		clear_commit_marks(list[i], assign_flag);
>> -	}
>> +	clear_commit_marks_many(nr_commits, list, RESULT | assign_flag);
>>   	free(list);
>>   
>>   	for (i = 0; i < from->nr; i++)
>>
>> base-commit: 4067a64672f9db8ca38d5a2682a7cdba7938c18b
> This change looks good to me.
>
> This is a tangent, but while re-reading clear_commit_marks() and its
> helpers to refresh my memory, I found that the bottom-most helper in
> the callchain was written in a very confusing way, but it is not a
> fault of this clean-up.  I however suspect that it would not help us
> all that much to use clear_commit_marks_many() with its current
> implementation.  It first clears all commits on the first-parent
> chain from each list[] element, while accumulating the parent
> commits that are yet to be processed in a commit_list in LIFO order,
> and then consumes these accumulated side parents the same way.  We
> probably would benefit by rewriting clear_commit_marks_many() to
> traverse from all the tips given in list[] taking advantage of the
> generation numbers, using a prio queue to manage the commits
> yet-to-be-cleared, or something.

Another commit walk that could be improved by generation numbers? It's 
like my bat-signal!

Thanks for pointing me in that direction. I'll take a look.

-Stolee