Re: [PATCH v2] ls-files: filter pathspec before lstat

[Junio C Hamano <gitster@pobox.com>]

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

> Please make sure that your v2 is a response to v1; otherwise loses
> sight of the previous iteration.
>
>> Changes in v2:
>> - Restrict early matching to one pathspec, avoiding the regression Jeff
>>   demonstrated with many pathspecs.
>> - Add all-matching and many-pathspec performance results.
>> - Drop the Assisted-by trailer.
>> - Link to v1: https://patch.msgid.link/20260607-ls-files-pathspec-lstat-v1-1-8cf40b730146@gmail.com
>
> And it is *not* a replacement to force human to follow such a link.
>
> Instead, please make sure each piece of your e-mail identifies where
> it fits in the discussion thread by pointing the message of the
> previous round with its In-Reply-To: header.

I won't complain about them individually, but it seems that all the
other v2 in different topics from you share the same problem.

Documentation/SubmittingPatches expect that the messages on the same
topic are threaded with In-Reply-To: headers; e-mail based workflow
tools like "b4" offer a useful feature that lets the user to feed
the message ID of an earlier round and fetch the messages in the
latest round.  As the message IDs of an earlier round that have
become commits for v1 are known in the refs/notes/amlog notes
(published at the usual places), replacing a topic with its newer
iteration becomes:

 (0) check out the previous round.

 (1) learn the message ID of the previous round we have checked out
     using notes/amlog (e.g., "git show -s --notes=amlog HEAD"),

 (2) detach the HEAD at the base of the previous round (roughly "git
     checkout master...HEAD", but not always),

 (3) ask "b4 am" to fetch the latest round of the thread the message
     we found in step (1) belongs to, and apply these new patches,

 (4) run "git range-diff @{-1}...HEAD".

which is very much automatable.

Unless an author breaks the thread, that is.