Re: A test for replay of regression tests

On Mon, Jan 17, 2022 at 6:53 PM Noah Misch <noah(at)leadboat(dot)com> wrote:
> On Mon, Jan 17, 2022 at 05:25:19PM +1300, Thomas Munro wrote:
> > Here's how it failed on fairywren, in case someone knowledgeable of
> > MSYS path translation etc can spot the problem:

> You likely need a PostgreSQL::Test::Utils::perl2host() call. MSYS Perl
> understands Cygwin-style names like /home/... as well as Windows-style names,
> but this PostgreSQL configuration understands only Windows-style names.

Thanks. I added that and pushed. Let's see if fairywren likes it
when it comes back online.

I also learned that in the CI environment, node->basedir() is a path
containing an internal "." component (I mean "something/./something").
I added a regex to collapse those, because they're unacceptable in
Windows junction point targets. I'm aware that there is something
happening in another CF entry that might address that sort of
thing[1], so then perhaps I could remove the kludge.

I tested that with a throw-away change to .cirrus.yml, like so. The
CI thread[2] is discussing a proper solution to these Windows CI blind
spots.

test_modules_script:
- perl src/tools/msvc/vcregress.pl modulescheck
+ - perl src/tools/msvc/vcregress.pl taptest ./src/test/modules/test_misc

[1] https://commitfest.postgresql.org/36/3331/
[2] https://www.postgresql.org/message-id/20220114235457.GQ14051%40telsasoft.com