Re: [HACKERS] Open 6.5 items

Tom Lane wrote:
>
> Vadim Mikheev <vadim(at)krs(dot)ru> writes:
> >> ERROR: Child itemid marked as unused
> > [ is fixed ]
>
> Great! Vadim (also Hiroshi and Tatsuo), how many bugs remain on your
> must-fix-for-6.5 lists? I was just wondering over in the "Freezing
> docs" thread whether we had any problems severe enough to justify
> delaying the release. It sounds like at least one such problem is
> gone...

No one in mine.

There are still some bad things, but they are old:

1. elog(NOTICE) in lock manager when locking was not succeeded.
Hope that our recent changes will reduce possibility of this.

Hiroshi wrote:

2.
> spinlock io_in_progress_lock of a buffer page is not
> released by operations called by elog() such as
> ProcReleaseSpins(),ResetBufferPool() etc.

I tried to fix this before 6.4 but without success (don't
remember why).

3.
> It seems elog(FATAL) doesn't release allocated buffer pages.
> It's OK ?
> AFAIC elog(FATAL) causes proc_exit(0) and proc_exit() doesn't
> call ResetBufferPool().

Seems to me that elog(FATAL) should call siglongjmp(Warn_restart, 1),
like elog(ERROR), but force exit in tcop main loop after
AbortCurrentTransaction(). AbortTransaction() does pretty nice
things like RelationPurgeLocalRelation(false) and DestroyNoNameRels()...

Vadim