Problem to "current-status information in shared memory" patch

Started by Qingqing Zhouover 19 years ago4 messages
#1Qingqing Zhou
zhouqq@cs.toronto.edu

On CVS tip:

$ postgres --single -Ddata postgres

PostgreSQL stand-alone backend 8.2devel
Segmentation fault (core dumped)

$ gdb -c core.18119
(gdb) bt
#0 pgstat_report_activity (cmd_str=0x67d25e "<IDLE>") at pgstat.c:1484
1484 beentry->st_changecount++;
(gdb) bt
#0 pgstat_report_activity (cmd_str=0x67d25e "<IDLE>") at pgstat.c:1484
#1 0x000000000056db8c in PostgresMain (argc=1, argv=0x8657d8,
username=0x867030 "qqzhou") at postgres.c:3132
#2 0x000000000050c918 in main (argc=4, argv=0x8657d0) at main.c:372
(gdb) p beentry
$1 = (volatile PgBackendStatus *) 0x0

Ooops, guess we forget to handle the single user mode in "current-status
information in shared memory" patch ...

Regards,
Qingqing

#2Tom Lane
tgl@sss.pgh.pa.us
In reply to: Qingqing Zhou (#1)
Re: Problem to "current-status information in shared memory" patch

"Qingqing Zhou" <zhouqq@cs.toronto.edu> writes:

Ooops, guess we forget to handle the single user mode in "current-status
information in shared memory" patch ...

"postgres --single" works for me. Maybe you need a "make
distclean"/rebuild?

regards, tom lane

#3Qingqing Zhou
zhouqq@cs.toronto.edu
In reply to: Qingqing Zhou (#1)
Re: Problem to "current-status information in shared memory" patch

"Tom Lane" <tgl@sss.pgh.pa.us> wrote

"postgres --single" works for me. Maybe you need a "make
distclean"/rebuild?

Sorry, because I forget to say that you need to turn "stats_command_string"
on. When it is off, pgstat_report_activity() will return immediately so
nothing wrong could be observed.

Regards,
Qingqing

#4Tom Lane
tgl@sss.pgh.pa.us
In reply to: Qingqing Zhou (#3)
Re: Problem to "current-status information in shared memory" patch

"Qingqing Zhou" <zhouqq@cs.toronto.edu> writes:

"Tom Lane" <tgl@sss.pgh.pa.us> wrote

"postgres --single" works for me. Maybe you need a "make
distclean"/rebuild?

Sorry, because I forget to say that you need to turn "stats_command_string"
on.

Ah. OK, fixed.

regards, tom lane