Horton

Re: longfin and tamandua aren't too happy but I'm not sure why

Thomas Munro <thomas.munro@gmail.com>

View thread View on postgr.es
From: Thomas Munro <thomas.munro@gmail.com>
To: Dilip Kumar <dilipbalaut@gmail.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Robert Haas <robertmhaas@gmail.com>, pgsql-hackers@lists.postgresql.org, Justin Pryzby <pryzby@telsasoft.com>
Date: 2022-09-28T10:35:16Z
Lists: pgsql-hackers 
On Wed, Sep 28, 2022 at 9:26 PM Dilip Kumar <dilipbalaut@gmail.com> wrote:
> It was a silly mistake, I used the F_OIDEQ function instead of
> F_INT8EQ. Although this was correct on the 0003 patch where we have
> removed the tablespace from key, but got missed in this :(
>
> I have locally reproduced this in a 32 bit machine consistently and
> the attached patch is fixing the issue for me.

I tested this with an armhf (32 bit) toolchain, and it passes
check-world, and was failing before.

Robert's patch isn't needed on this system. I didn't look into this
subject for long but it seems that SIGBUS on misaligned access (as
typically seen on eg SPARC) requires a 32 bit Linux/ARM kernel, but I
was testing with 32 bit processes and a 64 bit kernel.  Apparently 32
bit Linux/ARM has a control /proc/cpu/alignment to select behaviour
(options include emulation, SIGBUS) but 64 bit kernels don't have it
and are happy with misaligned access.

Commits

  1. meson: Split 'main' suite into 'regress' and 'isolation'

  2. ci: Add 32bit build and test

  3. Fix InitializeRelfilenumberMap for 05d4cbf9b6ba708858984b01ca0fc56d59d4ec7c

  4. Fix alignment problems with SharedInvalSmgrMsg.

  5. In BufTagGetForkNum, cast to the correct type.

  6. Increase width of RelFileNumbers from 32 bits to 56 bits.

  7. Harden pg_filenode_relation test against concurrent DROP TABLE.