> On 1 Dec 2021, at 12:49, Daniel Gustafsson <[email protected]> wrote: > >> On 1 Dec 2021, at 07:19, Michael Paquier <[email protected]> wrote: >> >> On Tue, Nov 30, 2021 at 11:34:21PM -0500, Tom Lane wrote: >>> Daniel Gustafsson <[email protected]> writes: >>>> Oh, interesting, I hadn't realized that. I'll go do that instead then. >>> >>> ... okay, so all we learned is that it really is an ENOENT failure. >>> >>> At this point my guess is that the test is copying the key file >>> to the wrong place because of an MSys path issue. I don't know >>> that topic well enough to debug it, though. >> >> Daniel has pinged me about this issue. From what I can see, c113d8ad >> has changed 001_ssltests.pl so as the keys are not anymore relative >> paths with *_tmp* names but absolute paths with the same key name, so >> it seems to me that you should sprinkle some perl2host() calls in the >> ${PostgreSQL::Test::Utils::tmp_check} paths to allow Msys to >> understand them. At quick glance, it looks like this is the first >> time we'd pass down absolute paths to libpq within connection >> strings in the TAP tests. > > I think that's a very plausible explanation, I'm currently experimenting with > a > patch that I will soon apply hoping that it will remedy the msys issue. If it > doesn't, I'll revert to copying inside ssl/ as before to return to the > drawingboard for a proper fix.
Looks like perl2host() was the missing piece, fairywren turned green with commit c3b34a0ff4. Thanks for pointing me in the right direction, I will draft a small paragraph on this to the TAP test README for other to learn from. -- Daniel Gustafsson https://vmware.com/
