Re: [PATCH v7 23/28] Reftable support for git-core

From: Han-Wen Nienhuys <hanwen@google.com>
Date: May 18, 2021, 13:30 UTC
Message-ID: <CAFQ2z_Mz65QsFEsdAEKrRkA8LpOcf+i_Yq-_Rb5sRj4GT=5h5A@mail.gmail.com>
In-Reply-To: <CAFQ2z_Os02GHCcPX0Hcin7+va232Di5wFBbdbOBqEehJvTFbFA@mail.gmail.com>

On Tue, May 18, 2021 at 3:18 PM Han-Wen Nienhuys <hanwen@google.com> wrote:

Show 13 quoted lines

> > This seems to assume that 'ref.value_type == REFTABLE_REF_VAL1' - but do
> > we expect to have to handle the other types
> > (REFTABLE_REF_VAL2/REFTABLE_REF_SYMREF)? When I run tests in seen
> > against ASAN I see the following errors in t0031, which suggests we're
> > running this code against REFTABLE_REF_SYMREF too - but I don't know if
> > that means that this code should be able to handle the other ref types
> > or if there's a bug higher up the stack. (AFAIUI REFTABLE_REF_DELETION
> > is already handled because reftable_table_read_ref() already returns 1
> > for deletion, but the other cases seem valid?)
>
> Curious.  I didn't know it was supported to create symrefs in a
> transaction, but I've added an assert to make sure this doesn't
> trigger any sanitizers.

This happens on repo initialization, where we create HEAD and its referents at the same time.

I've fixed this provisionally for now in my series.

-- 
Han-Wen Nienhuys - Google Munich
I work 80%. Don't expect answers from me on Fridays.
--

Google Germany GmbH, Erika-Mann-Strasse 33, 80636 Munich

Registergericht und -nummer: Hamburg, HRB 86891

Sitz der Gesellschaft: Hamburg

Geschäftsführer: Paul Manicle, Halimah DeLaine Prado