Re: [PATCH/RFC 6/7] refs: add update_refs for multiple simultaneous updates

From: Junio C Hamano <gitster@pobox.com>
Date: Aug 29, 2013, 18:32 UTC
Message-ID: <xmqqmwo04ac2.fsf@gitster.dls.corp.google.com>
In-Reply-To: <521F90EC.6040208@kitware.com>

Brad King <brad.king@kitware.com> writes:

Show 15 quoted lines

> On 08/29/2013 01:39 PM, Junio C Hamano wrote:
>> Brad King <brad.king@kitware.com> writes:
>>> +	for (i=0; i < n; ++i) {
>> 
>> Style:
>> 
>> 	for (i = 0; i < n; i++) {
>
> Fixed.
>
>> Is it asking for AB-BA deadlock?  If so, is the caller responsible
>> for avoiding it?
>
> Since we don't actually block waiting for locks we won't really
> deadlock.

Ahh, OK.

> For Git's internal API I think we can document this in a comment so
> that update_refs does not have to sort.  Then we can add a new
> ref_update_sort function to sort an array of struct ref_update.
> The user-facing "update-ref --stdin" can then use ref_update_sort.

My immediate reaction was "is there a case where the caller knows that it already has a sorted collection?". The single caller you are envisioning could collect the proposed updates to a string list and dedup, I think, and the resulting list would then be already sorted.

But it may not be a bad idea to keep the callers dumb and have this function always sort, dedup, *and* fail inconsistent request. Then your original caller that just collects --stdin input can pass possibly unsorted, duplicated and/or inconsistent request to the function and have it do the sanity check.