Justin Pryzby <pryzby@telsasoft.com> writes:
> On Fri, Feb 11, 2022 at 09:59:55AM -0500, Tom Lane wrote:
>> This seems like a pretty bad idea even if it weren't failing outright.
>> We should be examining the version of the file that's in the source
>> tree; the one in the installation tree might have version-skew
>> problems, if you've not yet done "make install".
> My original way used the source tree, but Michael thought it would be an issue
> for "installcheck" where the config may not be available.
Yeah, you are at risk either way, but in practice nobody is going to be
running these TAP tests without a source tree.
> This is what I had written:
> FROM (SELECT regexp_split_to_table(pg_read_file('postgresql.conf'), '\n') AS ln) conf
That's not using the source tree either, but the copy in the
cluster-under-test. I'd fear it to be unstable in the buildfarm, where
animals can append whatever they please to the config file being used by
tests.
regards, tom lane