Bug #110979 for DBD-Firebird: Compiler-dependent failure on RETURNING $sth->fetch

Thu Jan 07 08:37:25 2016 ribasushi [...] leporine.io - Ticket created

Subject:

Compiler-dependent failure on RETURNING $sth->fetch

I need to start with a disclaimer: this bug is reliably reproducible on travis on 5.8. However with a newer compiler stack (gcc 5.x) I am able to reproduce it even on perl 5.16 on my local workstation. So please do not dismiss this lightly. The setup is really simple - a plain insert-returning with a subsequent fetch, followed by a subsequent finish as seen in [1] fails with an exception [2] IFF we are running on a sufficiently new version of gcc (4.8+). This problem is reminiscent of the (yet unsolved) nonnull clusterfuck within perl itself as seen in [3]. I have not dug into the source to try to figure this out at all. The entire thing is reproducible with the steps at [4]. Cheers! [1] https://github.com/ribasushi/dbd_firebird_finish_failure/blob/master/test_firebird.t#L50-L58 [2] https://travis-ci.org/ribasushi/dbd_firebird_finish_failure/jobs/100819448#L1117 [3] https://rt.perl.org/Public/Bug/Display.html?id=125570 [4] https://github.com/ribasushi/dbd_firebird_finish_failure

Sat Jan 30 14:45:38 2016 dam [...] cpan.org - Correspondence added

На 07 ян. 2016, чт 15:37:25, RIBASUSHI написа: Show quoted text

> I need to start with a disclaimer: this bug is reliably reproducible > on travis on 5.8. However with a newer compiler stack (gcc 5.x) I am > able to reproduce it even on perl 5.16 on my local workstation. So > please do not dismiss this lightly. > > The setup is really simple - a plain insert-returning with a > subsequent fetch, followed by a subsequent finish as seen in [1] fails > with an exception [2] IFF we are running on a sufficiently new version > of gcc (4.8+). This problem is reminiscent of the (yet unsolved) > nonnull clusterfuck within perl itself as seen in [3]. > > I have not dug into the source to try to figure this out at all. The > entire thing is reproducible with the steps at [4]. > > Cheers! > > [1] > https://github.com/ribasushi/dbd_firebird_finish_failure/blob/master/test_firebird.t#L50- > L58 > [2] https://travis- > ci.org/ribasushi/dbd_firebird_finish_failure/jobs/100819448#L1117 > [3] https://rt.perl.org/Public/Bug/Display.html?id=125570 > [4] https://github.com/ribasushi/dbd_firebird_finish_failure

I can't reproduce this with the test in https://github.com/mariuz/perl-dbd-firebird/blob/master/t/rt110979.t (using gcc 5.3.1 and perl 5.22.1). Perhaps the CPAN testers will sched some light.

Sat Jan 30 14:45:38 2016 The RT System itself - Status changed from 'new' to 'open'

Sat Dec 02 05:12:15 2017 dam [...] cpan.org - Correspondence added

На 07 ян. 2016, чт 15:37:25, RIBASUSHI написа: Show quoted text

> I need to start with a disclaimer: this bug is reliably reproducible > on travis on 5.8. However with a newer compiler stack (gcc 5.x) I am > able to reproduce it even on perl 5.16 on my local workstation. So > please do not dismiss this lightly.

Sorry it took so long. the problem was with spurious error checking of uninitialised "status vector" (the thing Firebird uses to pass error conditions). Fixed in 1.27

Sat Dec 02 05:12:16 2017 dam [...] cpan.org - Status changed from 'open' to 'resolved'

Sat Dec 02 05:12:16 2017 dam [...] cpan.org - Fixed in 1.27 added

Fri Feb 23 08:36:24 2018 ribasushi [...] leporine.io - Correspondence added

On Sat Dec 02 11:12:15 2017, DAM wrote: Show quoted text

> > Fixed in 1.27

Confirmed: https://travis-ci.org/ribasushi/dbd_firebird_finish_failure/builds/100819447