On Fri, Oct 14, 2022 at 10:24:41AM +0200, Alvaro Herrera wrote:
> However, what's most problematic about this patch is that it introduces
> a pretty serious bug, yet that bug goes unnoticed if you just run the
> builtin test suites. I only noticed because I added an elog(ERROR,
> "oops") in the area protected by ENSURE_ERROR_CLEANUP and a debug
> elog(WARNING) in the cleanup area, then examined the server log after
> the pg_basebackup test filed; but this is not very workable. I wonder
> what would be a good way to keep this in check. The naive way seems to
> be to run a pg_basebackup, have it abort partway through (how?), then
> test the server and see if forced page writes are enabled or not.
See around the bottom of 010_pg_basebackup.pl, where a combination of
IPC::Run::start('pg_basebackup') with --max-rate and
pg_terminate_backend() is able to achieve that.
--
Michael
