Re: BUG #6065: FATAL: lock 0 not held

"Ben" <bwtest24@yahoo.com> writes:
> Operating system:   üLinux version 2.6.21-uc0 on ARM processor
> (NXP-LPC2478)
> Description:        FATAL: lock 0 not held
> Details:

> while initialising with initdb, after creating and filling new WAL file.
> We see LWLockAcquire(11) followed by LWLockRelease(0) and then a FATAL:
> lock 0 is not held.

Compiler bug?  It's pretty hard to credit that you've found some new
code path through bootstrap where the wrong lock number gets frobbed.

We do have at least one buildfarm member running on ARM, so that
architecture does get tested regularly, which eliminates various
other theories about PG portability problems.

            regards, tom lane