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

Hi,

While working on another round of the long option and fat comma style
cleanup, I noticed that the test for pg_upgrade --set-char-signedess
doesn't test what it's supposed to:

Masahiko Sawada <sawada(dot)mshk(at)gmail(dot)com> writes:

> diff --git a/src/bin/pg_upgrade/t/005_char_signedness.pl b/src/bin/pg_upgrade/t/005_char_signedness.pl
> index 05c3014a27d..c024106863e 100644
> --- a/src/bin/pg_upgrade/t/005_char_signedness.pl
> +++ b/src/bin/pg_upgrade/t/005_char_signedness.pl
> @@ -40,6 +40,23 @@ command_like(
> qr/Default char data signedness:\s+unsigned/,
> 'updated default char signedness is unsigned in control file');
>
> +# Cannot use --set-char-signedness option for upgrading from v18+
> +command_fails(
> + [
> + 'pg_upgrade', '--no-sync',
> + '-d', $old->data_dir,
> + '-D', $new->data_dir,
> + '-b', $old->config_data('--bindir'),
> + '-B', $new->config_data('--bindir'),
> + '-s', $new->host,
> + '-p', $old->port,
> + '-P', $new->port,
> + '-set-char-signedness', 'signed',

This is missing a dash, which causes the command to fail, but for the
wrong reason. pg_uprade seems to print all its errors on stdout, which
I guess is why the test use plain command_fails() instead of
command_fails_like(). However, we have another function to deal with
this: command_checks_all(). Attached are patches that fix the above
test, and also convert the other command_fails() calls in the pg_upgrade
tests to test for specific messages.

- ilmari