Re: [PATCH] gc --auto: do not run 'pack-refs' and 'reflog expire' twice

From: Junio C Hamano <gitster@pobox.com>
Date: Sep 2, 2014, 19:16 UTC
Message-ID: <xmqqtx4pzvcp.fsf@gitster.dls.corp.google.com>
In-Reply-To: <xmqqzjeh28p0.fsf@gitster.dls.corp.google.com>

Junio C Hamano <gitster@pobox.com> writes:

Show 8 quoted lines

> Nguyễn Thái Ngọc Duy  <pclouds@gmail.com> writes:
>
>> In the --auto code path, gc_before_repack() is called once in parent
>> process then again in the forked process. Stop the second run.
>
> Hmph.  Is the true reason why this happens because we have two calls
> to gc_before_repack() in the main, and one is before daemonize()?
> Shouldn't this protection be in the caller?

Actually, I think the function already is protecting wastage of getting called twice by saying "I've done pack-refs already" and "I've done prune-reflogs already" by setting the two variables to zero. Isn't that sufficient?

What problem is this patch trying to fix?

Show 20 quoted lines

>> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
>> ---
>>  builtin/gc.c | 5 +++++
>>  1 file changed, 5 insertions(+)
>>
>> diff --git a/builtin/gc.c b/builtin/gc.c
>> index 8d219d8..fb0cff3 100644
>> --- a/builtin/gc.c
>> +++ b/builtin/gc.c
>> @@ -261,6 +261,11 @@ static const char *lock_repo_for_gc(int force, pid_t* ret_pid)
>>  
>>  static int gc_before_repack(void)
>>  {
>> +	static int nr_runs;
>> +
>> +	if (nr_runs++)
>> +		return 0;
>> +
>>  	if (pack_refs && run_command_v_opt(pack_refs_cmd.argv, RUN_GIT_CMD))
>>  		return error(FAILED_RUN, pack_refs_cmd.argv[0]);