Обсуждение: assertion failures on BuildFarm that happened in slab.c

Hi, hackers


I've found two assertion failures on BuildFarm [1][2].
The call stack can be found in [2].

TRAP: failed Assert("dlist_is_empty(blocklist)"), File: "slab.c", Line: 564, PID: 16148
postgres: publisher: walsender postgres postgres [local] START_REPLICATION(ExceptionalCondition+0x54)[0x983564]
postgres: publisher: walsender postgres postgres [local] START_REPLICATION(SlabAlloc+0x50d)[0x9b554d]
postgres: publisher: walsender postgres postgres [local] START_REPLICATION(MemoryContextAlloc+0x4c)[0x9b2b2c]
postgres: publisher: walsender postgres postgres [local] START_REPLICATION(ReorderBufferGetChange+0x15)[0x7d62d5]
postgres: publisher: walsender postgres postgres [local] START_REPLICATION(heap_decode+0x4d1)[0x7cb071]
postgres: publisher: walsender postgres postgres [local] START_REPLICATION(LogicalDecodingProcessRecord+0x63)[0x7ca033]
postgres: publisher: walsender postgres postgres [local] START_REPLICATION[0x7f1962]
postgres: publisher: walsender postgres postgres [local] START_REPLICATION[0x7f42e2]
postgres: publisher: walsender postgres postgres [local] START_REPLICATION(exec_replication_command+0xa88)[0x7f4ec8]
postgres: publisher: walsender postgres postgres [local] START_REPLICATION(PostgresMain+0x524)[0x847a44]
2022-12-20 13:18:39.725 UTC [16167:4] 015_stream.pl LOG:  disconnection: session time: 0:00:00.019 user=postgres
database=postgreshost=[local] 
postgres: publisher: walsender postgres postgres [local] START_REPLICATION[0x7b57e6]
postgres: publisher: walsender postgres postgres [local] START_REPLICATION(PostmasterMain+0xe27)[0x7b6757]
postgres: publisher: walsender postgres postgres [local] START_REPLICATION(main+0x683)[0x4c74e3]
/lib/libc.so.6(__libc_start_main+0xf0)[0x7f5179a8d070]
postgres: publisher: walsender postgres postgres [local] START_REPLICATION(_start+0x2a)[0x4c75aa]
2022-12-20 13:18:39.795 UTC [16107:4] LOG:  server process (PID 16148) was terminated by signal 6: Aborted


The last failure for subscriptionCheck before those for HEAD happened 66 days ago [3]
and I checked all failures there within 90 days. There is no similar failure.

I'm not sure, but this could be related to a recent commit (d21ded75fd) in [4].

[1] - https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=butterflyfish&dt=2022-12-20%2013%3A00%3A19
[2] - https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=grassquit&dt=2022-12-20%2010%3A34%3A39
[3] - https://buildfarm.postgresql.org/cgi-bin/show_failures.pl?max_days=90&stage=subscriptionCheck&filter=Submit
[4] - https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=d21ded75fdbc18d68be6e6c172f0f842c50e9263



Best Regards,
    Takamichi Osumi

On Wed, Dec 21, 2022 at 03:28:01AM +0000, Takamichi Osumi (Fujitsu) wrote:
> The last failure for subscriptionCheck before those for HEAD happened 66 days ago [3]
> and I checked all failures there within 90 days. There is no similar failure.
>
> I'm not sure, but this could be related to a recent commit (d21ded75fd) in [4].
>
> [1] - https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=butterflyfish&dt=2022-12-20%2013%3A00%3A19
> [2] - https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=grassquit&dt=2022-12-20%2010%3A34%3A39
> [3] - https://buildfarm.postgresql.org/cgi-bin/show_failures.pl?max_days=90&stage=subscriptionCheck&filter=Submit
> [4] - https://git.postgresql.org/gitweb/?p=postgresql.git;a=commitdiff;h=d21ded75fdbc18d68be6e6c172f0f842c50e9263

Good catch.  Yes, d21ded7 looks to have issues.  I am adding David in
CC.
--
Michael

On Wed, 21 Dec 2022 at 18:04, Michael Paquier <michael@paquier.xyz> wrote:
> Good catch.  Yes, d21ded7 looks to have issues.  I am adding David in
> CC.

Thanks. I'll look now.

David

On Wed, 21 Dec 2022 at 16:28, Takamichi Osumi (Fujitsu)
<osumi.takamichi@fujitsu.com> wrote:
> TRAP: failed Assert("dlist_is_empty(blocklist)"), File: "slab.c", Line: 564, PID: 16148

> I'm not sure, but this could be related to a recent commit (d21ded75fd) in [4].

It was. I've just pushed a fix.  Thanks for highlighting the problem.

David