Re: [bug fix] Produce a crash dump before main() on Windows

On Wed, Jul 18, 2018 at 4:10 PM Tsunakawa, Takayuki <
tsunakawa(dot)takay(at)jp(dot)fujitsu(dot)com> wrote:

> From: Haribabu Kommi [mailto:kommi(dot)haribabu(at)gmail(dot)com]
> > I reviewed patch and it works as per the subject, but I am not able to
> verify
> > the actual
> > bug that is reported in the upthread. The moving of setErrorMode() call
> > to the start
> > of the main function can handle all the cases that can lead to a crash
> with
> > no popup.
>
> Thank you for reviewing the patch. Yeah, I don't know reproduce the
> problem too, because some bug of Symantec AntiVirus randomly caused the
> crash before main()...
>
>
> > The reset of setErrorMode(0) before start of any process can generate the
> > coredump
> > because of any issue before it reaches the main() function, but this
> change
> > can lead
> > to stop the postmaster restart process, if no one present to observe the
> > scenario?
> > Doesn't this change can generate backward compatibility problems to
> > particular users?
>
> Sorry, I couldn't get it. Could you elaborate on it? I couldn't
> understand how this change has an adverse effect on the postmaster restart
> process (restart_after_crash = on) and backward compatibility.
>

Sorry about not providing more details. It is just my doubt, because
I am not able to reproduce the problem to test it after the fix. May be
I can give a try by modifying the source code to get the crash.

My point is, With this patch, in case if the postgres crashses
before reaching main(), does it generate a popup?

If the answer to the above question is yes, then if the popup is
not attended by the DBA, I feel the postmaster is not able to
restart all the processes because it is waiting for the crash process
to close. Am I Correct?

Regards,
Haribabu Kommi
Fujitsu Australia