Re: [PATCH] plpython function causes server panic

I wrote:
> Hao Zhang <zhrt1446384557(at)gmail(dot)com> writes:
>> IMHO, there are other error reports in the function
>> BeginInternalSubTransaction(), like

> Sure, but all the other ones are extremely hard to hit, which is why
> we didn't bother to worry about them to begin with. If we want to
> make this more formally bulletproof, my inclination would be to
> (a) get rid of the IsInParallelMode restriction and then (b) turn
> the function into a critical section, so that any other error gets
> treated as a PANIC.

Here's a draft patch along this line. Basically the idea is that
subtransactions used for error control are now legal in parallel
mode (including in parallel workers) so long as they don't try to
acquire their own XIDs. I had to clean up some error handling
in xact.c, but really this is a pretty simple patch.

Rather than a true critical section (ie PANIC on failure), it seemed
to me to be enough to force FATAL exit if BeginInternalSubTransaction
fails. Given the likelihood that our transaction state is messed up
if we get a failure partway through, it's not clear to me that we
could do much better than that even if we were willing to make an API
change for BeginInternalSubTransaction.

I haven't thought hard about what new test cases we might want to
add for this. It gets through check-world as is, meaning that
nobody has made any test cases exercising the previous restrictions
either. There might be more documentation work to be done, too.

regards, tom lane