Re: Re: Parallel scan with SubTransGetTopmostTransaction assert coredump

On Mon, May 24, 2021 at 11:56 PM Pavel Borisov <pashkin(dot)elfe(at)gmail(dot)com> wrote:
>
> Using a recipe similar to what has been described above in the thread, I reliably reproduced the bug in many Postgres versions. (v.11, v.13 etc.).
> 1. Make & make install
> 2. Make check
> 3. run SubTransGetTopmostTransaction-rep.sh in the Postgres source code dir.
>

Thanks, using the latest PG14 code
(713a431c781fbfe1a22fae4991836077f0f4c513), with no modifications, and
building with "--enable-cassert", your script reproduced the coredump
on my system in about 10 minutes, on the 1st run.

#0 0x00007f1542d1137f in raise () from /lib64/libc.so.6
#1 0x00007f1542cfbdb5 in abort () from /lib64/libc.so.6
#2 0x00000000009380cb in ExceptionalCondition ()
#3 0x00000000005aadbc in SubTransGetTopmostTransaction ()
#4 0x00000000009741ef in XidInMVCCSnapshot ()
#5 0x0000000000563326 in HeapTupleSatisfiesVisibility ()
#6 0x0000000000559d50 in heapgetpage ()
#7 0x000000000055a0e9 in heapgettup_pagemode ()
#8 0x000000000055b98e in heap_getnextslot ()
#9 0x00000000006df713 in SeqNext ()
#10 0x00000000006b725f in ExecScan ()
#11 0x00000000006bef41 in fetch_input_tuple ()
#12 0x00000000006c1f6f in ExecAgg ()
#13 0x00000000006ade83 in standard_ExecutorRun ()
#14 0x00000000006b21ed in ParallelQueryMain ()
#15 0x00000000005a8c11 in ParallelWorkerMain ()
#16 0x0000000000795f3c in StartBackgroundWorker ()
#17 0x00000000007a1bc1 in maybe_start_bgworkers ()
#18 0x00000000007a25c5 in sigusr1_handler ()
#19 <signal handler called>
#20 0x00007f1542dce25b in select () from /lib64/libc.so.6
#21 0x00000000007a2cf6 in ServerLoop ()
#22 0x00000000007a4659 in PostmasterMain ()
#23 0x00000000005042ed in main ()

(Using a debug-build, it seems much harder to reproduce on my system)

With my patch applied, I've had no crashes, from many runs.

Regards,
Greg Nancarrow
Fujitsu Australia