On Sun, Jun 30, 2024 at 09:32:57PM -0400, Tom Lane wrote:
> Noah Misch <noah@leadboat.com> writes:
> > On Sun, Jun 30, 2024 at 05:01:52PM -0400, Tom Lane wrote:
> >> I'm tempted to lobotomize the new test case on Windows until
> >> we have that resolved.
>
> > Sounds fine. The pg_upgrade suite adequately tests appendShellString() and
> > appendConnStrVal() with the larger character repertoire, so it's enough for
> > pg_createsubscriber to test just a space or so.
>
> Hmph ... according to [1], 545082091 was not enough to fix this.
> I guess that old version of IPC::Run also misbehaves for cases
> involving backslash, single quote, and/or some other special
> characters?
The database name has backslash toward the end, which likely creates a
problematic backslash-double-quote sequence under win32 command line rules.
You can see that in the node log, comparing the CREATE DATABASE log line
(shows dquote at end of dbname) to the FATAL log line (no dquote at end). I'm
not aware of trouble with characters other than backslash or dquote.
> [1] https://buildfarm.postgresql.org/cgi-bin/show_log.pl?nm=fairywren&dt=2024-07-01%2000%3A03%3A05
