Re: Stack overflow issue

Hi,

On 2024-04-16 15:45:42 +0300, Alexander Korotkov wrote:
> On Tue, Apr 16, 2024 at 1:48 AM Andres Freund <andres(at)anarazel(dot)de> wrote:
> > On 2024-03-06 14:17:23 +0200, Alexander Korotkov wrote:
> > > 0001 Turn tail recursion into iteration in CommitTransactionCommand()
> > > I did minor revision of comments and code blocks order to improve the
> > > readability.
> >
> > After sending
> > https://www.postgresql.org/message-id/20240414223305.m3i5eju6zylabvln%40awork3.anarazel.de
> > I looked some more at important areas where changes didn't have code
> > coverage. One thing I noticed was that the "non-internal" part of
> > AbortCurrentTransaction() is uncovered:
> > https://anarazel.de/postgres/cov/16-vs-HEAD-2024-04-14/src/backend/access/transam/xact.c.gcov.html#L3403
> >
> > Which made me try to understand fefd9a3fed2. I'm a bit confused about why
> > some parts are handled in CommitCurrentTransaction()/AbortCurrentTransaction()
> > and others are in the *Internal functions.
> >
> > I understand that fefd9a3fed2 needed to remove the recursion in
> > CommitTransactionCommand()/AbortCurrentTransaction(). But I don't understand
> > why that means having some code in in the non-internal and some in the
> > internal functions? Wouldn't it be easier to just have all the state handling
> > code in the Internal() function and just break after the
> > CleanupSubTransaction() calls?
>
> I'm not sure I correctly get what you mean. Do you think the attached
> patch matches the direction you're pointing? The patch itself is not
> final, it requires cleanup and comments revision, just to check the
> direction.

Something like that, yea. The split does seem less confusing that way to me,
but also not 100% certain.

Greetings,

Andres Freund