On Tue, Mar 16, 2021 at 11:22 AM Michael Paquier <mich...@paquier.xyz> wrote: > > On Tue, Mar 16, 2021 at 10:02:25AM +0100, Magnus Hagander wrote: > > If you back out that patch, does the problem you can reproduce with > > archive_command go away? > > That's the first thing I did after seeing the failure, and I saw > nothing after 2~3 hours of pgbench :)
:) That's at least an "almost". > The second thing I did was to revert back to HEAD with more logging in > the area, but I was not able to see my error again. Perhaps I just > need to put more load, there are still too many guesses and not enough > facts. Agreed. > > I agree with your analysis in general. It certainly seems to hit right > > in the center of the problem scope. > > > > Maybe hardlinks on Windows has yet another "weird behaviour" vs what > > we're used to from Unix. > > Yeah, I'd like to think that this is a rational explanation, and > that's why I was just focusing on reproducing this issue rather > reliably as a first step. Yeah, it'd definitely be good to figure out exactly what it is that triggers the issue. > > It would definitely be more useful if we could figure out *when* this > > happens. But failing that, I wonder if we could find a way to provide > > a build with this patch backed out for the bug reporters to test out, > > given they all seem to have it fairly well reproducible. (But I am > > assuming are unlikely to be able to create their own builds easily, > > given the complexity of doing so on Windows). Given that this is a > > pretty isolated change, it should hopefully be easy enough to back out > > for testing. > > There is a large pool of bug reporters, hopefully one of them may be > able to help.. I think you're overestimating peoples ability to get our build going on Windows :) If we can provide a new .EXE built with exactly the same flags as the EDB downloads that they can just drop into a directory, I think it's a lot easier to get that done. -- Magnus Hagander Me: https://www.hagander.net/ Work: https://www.redpill-linpro.com/
