On Wed, Jul 20, 2022 at 1:28 PM Masahiko Sawada <sawada.mshk@gmail.com> wrote:
>
> On Wed, Jul 20, 2022 at 2:19 PM Amit Kapila <amit.kapila16@gmail.com> wrote:
> >
> > On Wed, Jul 20, 2022 at 9:01 AM Masahiko Sawada <sawada.mshk@gmail.com> wrote:
> >
> > > Another idea would be to have functions, say
> > > SnapBuildCommitTxnWithXInfo() and SnapBuildCommitTxn_ext(). The latter
> > > does actual work of handling transaction commits and both
> > > SnapBuildCommitTxn() and SnapBuildCommit() call
> > > SnapBuildCommitTxnWithXInfo() with different arguments.
> > >
> >
> > Do you want to say DecodeCommit() instead of SnapBuildCommit() in
> > above para?
>
> I meant that we will call like DecodeCommit() ->
> SnapBuildCommitTxnWithXInfo() -> SnapBuildCommitTxn_ext(has_invals =
> true) -> SnapBuildXidHasCatalogChanges(has_invals = true) -> ... If
> SnapBuildCommitTxn() gets called, it calls SnapBuildCommitTxn_ext()
> with has_invals = false and behaves the same as before.
>
Okay, understood. This will work.
> > Yet another idea could be to have another flag
> > RBTXN_HAS_INVALS which will be set by DecodeCommit for top-level TXN.
> > Then, we can retrieve it even for each of the subtxn's if and when
> > required.
>
> Do you mean that when checking if the subtransaction has catalog
> changes, we check if its top-level XID has this new flag?
>
Yes.
> Why do we
> need the new flag?
>
This is required if we don't want to introduce a new set of functions
as you proposed above. I am not sure which one is better w.r.t back
patching effort later but it seems to me using flag stuff would make
future back patches easier if we make any changes in
SnapBuildCommitTxn.
--
With Regards,
Amit Kapila.