On Fri, Dec 13, 2024 at 09:34:21PM -0500, Andres Freund wrote:
> On 2024-12-13 16:38:05 -0800, Noah Misch wrote:
> > On Fri, Dec 13, 2024 at 05:41:15PM -0500, Andres Freund wrote:
> > > Hm. Leaving RBM_ZERO_AND_LOCK aside, is it actually always safe to do
> > > RestoreBlockImage() into a buffer tha
Hi,
On 2024-12-14 13:41:40 +1300, Thomas Munro wrote:
> On Sat, Dec 14, 2024 at 11:41 AM Andres Freund wrote:
> > On 2024-05-12 10:16:58 -0700, Noah Misch wrote:
> > > The use of RBM_ZERO_AND_LOCK is incompatible with that. See a similar
> > > argument at https://postgr.es/m/flat/5101.1328219...
Hi,
On 2024-12-13 16:38:05 -0800, Noah Misch wrote:
> On Fri, Dec 13, 2024 at 05:41:15PM -0500, Andres Freund wrote:
> > Afaics we didn't do anything about this issue?
>
> postgr.es/c/e656657 fixed this. I thought this was longstanding, but it
> turned out to have started on 2024-04-02.
Ah, tha
On Sat, Dec 14, 2024 at 11:41 AM Andres Freund wrote:
> On 2024-05-12 10:16:58 -0700, Noah Misch wrote:
> > The use of RBM_ZERO_AND_LOCK is incompatible with that. See a similar
> > argument at https://postgr.es/m/flat/5101.1328219...@sss.pgh.pa.us that led
> > me to the cause. Adding a 10ms sle
On Fri, Dec 13, 2024 at 05:41:15PM -0500, Andres Freund wrote:
> Afaics we didn't do anything about this issue?
postgr.es/c/e656657 fixed this. I thought this was longstanding, but it
turned out to have started on 2024-04-02.
> Hm. Leaving RBM_ZERO_AND_LOCK aside, is it actually always safe to d
Hi,
Just found this thread because I was looking for discussions for some
behaviour of XLogReadBufferExtended()...
Afaics we didn't do anything about this issue?
I just tried to repro this, without success so far. But it very well might be
sufficiently timing dependent that it just happens to n
