Tatsuo Ishii <t-ishii@sra.co.jp> writes:
> > mklinux PPC750 7.0 2000-04-13, Tatsuo Ishii
> If compiled with -O2 or -O2 -g, I got 10 tests FAILED. misc test
> failed due to a backend crash. The SQL caused the crash was:
> select i, length(t), octet_length(t), oldstyle_length(i,t) from
> oldstyle_test;
> #0 ExecReplace (slot=0x1a4a7d0, tupleid=0x0, estate=0x1a4a708)
> at execMain.c:1408
> 1408 resultRelationDesc = resultRelInfo->ri_RelationDesc;
> (gdb) where
> #0 ExecReplace (slot=0x1a4a7d0, tupleid=0x0, estate=0x1a4a708)
> at execMain.c:1408
> #1 0x188471c in ExecutePlan (estate=0x0, plan=0x1a4a410,
> operation=CMD_SELECT, numberTuples=0, direction=27567836,
> destfunc=0x1a4adf8) at execMain.c:1127
> #2 0x188471c in ExecutePlan (estate=0x0, plan=0x1a4a410,
> operation=CMD_SELECT, numberTuples=0, direction=27567836,
> destfunc=0x1a4adf8) at execMain.c:1127
> #3 0x18838b8 in ExecutorRun (queryDesc=0x1a4a7d0, estate=0x1a4a708,
> feature=27567784, count=0) at execMain.c:233
I think you've got a badly broken compiler there. There's no way that
ExecReplace should be entered for a SELECT. The backtrace is wrong on
its face anyway --- ExecutePlan does not call itself.
What gcc version does that platform have?
regards, tom lane