Re: Walsender crash

On Mon, Feb 13, 2017 at 8:55 PM, Stas Kelvich <s(dot)kelvich(at)postgrespro(dot)ru> wrote:
> Hello.
>
> While playing with logical publications/subscriptions I’ve noted that walsender crashes
> when i’m trying to CREATE SUBSCRIPTION to a current server.
>
> So having one running postgres instance on a standard port:
>
> stas=# CREATE SUBSCRIPTION mysub CONNECTION 'dbname=stas host=127.0.0.1' PUBLICATION mypub;
> ^CCancel request sent
> ERROR: canceling statement due to user request
> stas=# CREATE SUBSCRIPTION mysub CONNECTION 'dbname=stas host=127.0.0.1' PUBLICATION mypub;
> WARNING: terminating connection because of crash of another server process
>
> Crash happens due to uninitialised StringBuffer reply_message in walsender.c:
>
> (lldb) bt
> * thread #1: tid = 0x13e3f, 0x000000010e74ebaf postgres`resetStringInfo(str=0x000000010ecd12a0) + 15 at stringinfo.c:96, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x0)
> * frame #0: 0x000000010e74ebaf postgres`resetStringInfo(str=0x000000010ecd12a0) + 15 at stringinfo.c:96
> frame #1: 0x000000010e8778c9 postgres`ProcessRepliesIfAny + 169 at walsender.c:1347
> frame #2: 0x000000010e8770fd postgres`WalSndWaitForWal(loc=35929928) + 221 at walsender.c:1142
> frame #3: 0x000000010e876cf2 postgres`logical_read_xlog_page(state=0x00007fbcac0446f8, targetPagePtr=35921920, reqLen=8008, targetRecPtr=35929904, cur_page="\x95?", pageTLI=0x00007fbcac044fa4) + 50 at walsender.c:717
> frame #4: 0x000000010e571e5e postgres`ReadPageInternal(state=0x00007fbcac0446f8, pageptr=35921920, reqLen=8008) + 542 at xlogreader.c:556
> frame #5: 0x000000010e571336 postgres`XLogReadRecord(state=0x00007fbcac0446f8, RecPtr=35929904, errormsg=0x00007fff5178f1e0) + 326 at xlogreader.c:255
> frame #6: 0x000000010e85ea9b postgres`DecodingContextFindStartpoint(ctx=0x00007fbcac044438) + 139 at logical.c:432
> frame #7: 0x000000010e874dfd postgres`CreateReplicationSlot(cmd=0x00007fbcad8004d0) + 333 at walsender.c:796
> frame #8: 0x000000010e8747a4 postgres`exec_replication_command(cmd_string="CREATE_REPLICATION_SLOT \"mysub\" LOGICAL pgoutput") + 532 at walsender.c:1272
> frame #9: 0x000000010e8e6adc postgres`PostgresMain(argc=1, argv=0x00007fbcad005f50, dbname="stas", username="stas") + 2332 at postgres.c:4064
> frame #10: 0x000000010e839abe postgres`BackendRun(port=0x00007fbcabc033a0) + 654 at postmaster.c:4310
> frame #11: 0x000000010e838eb3 postgres`BackendStartup(port=0x00007fbcabc033a0) + 483 at postmaster.c:3982
> frame #12: 0x000000010e837ea5 postgres`ServerLoop + 597 at postmaster.c:1722
> frame #13: 0x000000010e8356b5 postgres`PostmasterMain(argc=3, argv=0x00007fbcabc02b90) + 5397 at postmaster.c:1330
> frame #14: 0x000000010e76371f postgres`main(argc=3, argv=0x00007fbcabc02b90) + 751 at main.c:228
> frame #15: 0x00007fffa951c255 libdyld.dylib`start + 1
> frame #16: 0x00007fffa951c255 libdyld.dylib`start + 1
>
> Patch with lacking initStringInfo() attached.

Thanks for the report and patch!
I pushed the modified version of the patch.

Regards,

--
Fujii Masao