Horton

Re: pg_trgm comparison bug on cross-architecture replication due to different char implementation

Peter Eisentraut <peter@eisentraut.org>

View thread View on postgr.es
From: Peter Eisentraut <peter@eisentraut.org>
To: Masahiko Sawada <sawada.mshk@gmail.com>, Noah Misch <noah@leadboat.com>
Cc: Tom Lane <tgl@sss.pgh.pa.us>, Joe Conway <mail@joeconway.com>, Alexander Korotkov <aekorotkov@gmail.com>, "Guo, Adam" <adamguo@amazon.com>, "pgsql-hackers@lists.postgresql.org" <pgsql-hackers@lists.postgresql.org>, Nathan Bossart <nathandbossart@gmail.com>, Jim Mlodgenski <jimmy76@gmail.com>
Date: 2025-03-19T09:58:38Z
Lists: pgsql-hackers 
On 21.02.25 20:39, Masahiko Sawada wrote:
>>>> I have one question about the 0001 patch; since we add
>>>> 'default_char_signedness' field to ControlFileData do we need to bump
>>>> PG_CONTROL_VERSION? We have comments about bumping PG_CONTROL_VERSION
>>>> when changing CheckPoint struct or DBState enum so it seems likely but
>>>> I'd like to confirm just in case that we need to bump
>>>> PG_CONTROL_VERSION also when changing ControlFileData.
>>>
>>> Yes.  (I'm not aware of value we get from having distinct control file version
>>> and catalog version, but we do have both.)
>>>
>>>> If we need, can
>>>> we bump it to 1800? or 1701?
>>>
>>> I'd do 1800.  The pattern seems to be to bump to 1800 for the first pg_control
>>> change of the v18 cycle, then 1801, then 1802 for the third change of the
>>> cycle.  That's based on this history:
>>>
>>> git log -U0 -p src/include/catalog/pg_control.h | grep -E '^(Date|\+#define PG_CONTROL_VERSION)'
>>
>> Thank you for the confirmation. That makes sense to me.
>>
>> I'll push these patches with version bumps, barring any objections or
>> further comments.
> 
> Pushed.

Is there a reason why the pg_controldata and pg_resetwal output are 
"Default char *data* signedness", while the rest of the patch and 
description just says "char signedness"?  The word "data" doesn't mean 
anything here, does it?

Commits

  1. pg_upgrade: Check for the expected error message in TAP tests.

  2. Fix a typo in 005_char_signedness.pl test.

  3. Add test 005_char_signedness.pl to meson.build.

  4. Fix an issue with index scan using pg_trgm due to char signedness on different architectures.

  5. pg_upgrade: Add --set-char-signedness to set the default char signedness of new cluster.

  6. pg_upgrade: Preserve default char signedness value from old cluster.

  7. pg_resetwal: Add --char-signedness option to change the default char signedness.

  8. Add default_char_signedness field to ControlFileData.

  9. Remove unneeded nbtree array preprocessing assert.