Re: [PATCH] remote: introduce config to set prefetch refs

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Sun, Sep 15, 2024 at 9:42 PM Junio C Hamano <gitster@xxxxxxxxx> wrote:
>
> Junio C Hamano <gitster@xxxxxxxxx> writes:
>
> > Existing logic filteres them down to what matches configured
> > remote.$name.fetch variable.  filter_prefetch_refspec() may further
> > reduces the result by removing those whose .src side begins with
> > "refs/tags/".
> >
> > Now you only look at what survived the above existing filtering, and
> > further narrow it down by picking only ones that match the prefetch
> > condition.  If the refspec that survived the filtering by the fetch
> > refspec (and existing logic in filter_prefetch_refspec()) does not
> > satisfy the prefetch condition, it won't be prefetched.
>
> Sorry, but I misread the code.
>
> By the time filter_prefetch_refspec() is called by get_ref_map(),
> this caller has "remote_refs" linked list that describes each ref it
> is going to fetch, so conceptually what is left for the prefetch
> logic to do is to selectively discard the elements on this list that
> are not worth asking the remote to send new object data for and use
> the remainder of the list in remote_refs list, and the logic to
> further limit this list with the prefetchref configuration would fit
> well here, but filter_prefetch_refspec() does not work on this list
> at all X-<.  So the prefetchref limitation needs to come outside the
> function.

Sure, can you take a look at the latest diff (shared in the last reply)?
Moved the prefetchref filtering to `get_ref_map` where advertised refs
are available.





[Index of Archives]     [Linux Kernel Development]     [Gcc Help]     [IETF Annouce]     [DCCP]     [Netdev]     [Networking]     [Security]     [V4L]     [Bugtraq]     [Yosemite]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux SCSI]     [Fedora Users]

  Powered by Linux