Re: What's cooking in git.git (Sep 2021, #08; Mon, 27)

From: Elijah Newren <newren@gmail.com>
To: Junio C Hamano <gitster@pobox.com>
Cc: Jeff King <peff@peff.net>, Neeraj Singh <nksingh85@gmail.com>,
	Git Mailing List <git@vger.kernel.org>,
	Jonathan Nieder <jrnieder@gmail.com>
Subject: Re: What's cooking in git.git (Sep 2021, #08; Mon, 27)
Date: Fri, 1 Oct 2021 16:05:38 -0700	[thread overview]
Message-ID: <CABPp-BG9JNeFJ_gzepntcieJAMzT1OKNes9zEOiWY3JMXh4BMg@mail.gmail.com> (raw)
In-Reply-To: <xmqqr1d42ykj.fsf@gitster.g>

On Fri, Oct 1, 2021 at 3:02 PM Junio C Hamano <gitster@pobox.com> wrote:
>
> Elijah Newren <newren@gmail.com> writes:
>
> > So with that out of the way, let's return to discussing the textconv
> > cache.  If the remerge-diff results aren't cached, isn't it unsafe to
> > allow the textconv cache to persist anything while remerge-diff is
> > running because it could create corruption?
>
> I do not think anybody involved in this thread thinks it is
> practical to annotate each write_object_file() with "this is
> temporary" vs "this is to persist", so it is a given that it would
> be all-or-none.  If we want write_object_file() called while we are
> running remerge-diff to write to a temporary object store, we have
> to accept any other write_object_file() called by somebody else,
> like textconv cache, must become temporary.

Ok, good.

But is this write_object_file() specific?  I think
pretend_object_file() has the same problem where the textconv cache
could reference a pretend_object_file() and thus write objects and/or
refs that become corrupt.  (In particular, if the userdiff fires on
any of the new files created by the three-way content merges, then I
think we hit the exact same problem)  So, I think as long as there are
any "temporary" objects being fed to diff, the textconv cache needs to
also be considered temporary.  Or am I misunderstanding something?

> It may be sufficient to plug ref updates (which would cover the
> finialization of notes-cache used by the textconv cache) to avoid
> corruption, but that might give us a pointless and unpleasant error
> message, so it may be necessary to teach the notes-cache stuff to
> allow getting existing cached data while disabling it to accept
> cache updates.

All we need to do here is set the quarantine environment, as per
Neeraj's report[1], it already handles all of this:

"""
Peff's test case was invoking ref updates via the cachetextconv
setting. That particular code silently does nothing when a ref
update is forbidden
"""

Looking at the code, this appears to happen because
ref_transaction_prepare() doesn't print an error in the quarantine
environment, but returns one.  That error is passed all the way up the
stack to notes_cache_write() and then to fill_textconv().
fill_textconv() is the first level in the stack that ignores the
return code of what it's calling, namely notes_cache_write(), which
means that things work fine.  There's even a comment a few lines
stating, "ignore errors, as we might be in a readonly repository".
So, I think we're good as long as we ensure the quarantine environment
is setup.

[1] https://lore.kernel.org/git/20210929184339.GA19712@neerajsi-x1.localdomain/