Re: fairywren hung in pg_basebackup tests

On 2022-07-24 Su 15:10, Noah Misch wrote:
> On Sun, Jul 24, 2022 at 12:55:56PM -0400, Tom Lane wrote:
>> Andrew Dunstan <andrew(at)dunslane(dot)net> writes:
>>> fairywren (msys2 animal) is currently hung in the pg_basebackup tests.
>>> Here's the bottom of the regress log. I don't have further info as yet,
>>> but can dig is someone has a suggestion.
>> Hm, what's with the "Use of uninitialized value" warnings?
> The warnings are sequelae of:
>
>>> [13:11:33.593](0.000s) not ok 97 - one tablespace tar was created
> >From that, it follows that $tblspc_tars[0] is undef at:
>
> PostgreSQL::Test::Utils::system_or_bail($tar, 'xf', $tblspc_tars[0],
> '-C', $repTsDir);
>
>>> # Running: C:/Windows/System32/tar xf  -C
>>> C:\tools\nmsys64\home\pgrunner\bf\root\REL_15_STABLE\pgsql.build\src\bin\pg_basebackup\tmp_check\tmp_test_jVCb/tblspc1replica

Perhaps we should have a guard in system_or_bail() and/or system_log()
which bails if some element of @_ is undefined.

> I can confirm that Windows tar hangs when invoked that way. For preventing
> the hang, the test file could die() or skip the tar-program-using section
> after failing 'one tablespace tar was created'. That still leaves a question
> about why pg_basebackup didn't make the tablespace tar file. I would check
> 010_pg_basebackup_main.log for clues about that.

The same thing has happened again on HEAD. I don't see anything in that
file that gives any clue.

cheers

andrew

--
Andrew Dunstan
EDB: https://www.enterprisedb.com