Re: Built-in CTYPE provider

On Tue, Mar 19, 2024 at 11:55 AM Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote:
> Jeff Davis <pgsql(at)j-davis(dot)com> writes:
> > On Mon, 2024-03-18 at 18:04 -0400, Tom Lane wrote:
> >> This is causing all CI jobs to fail the "compiler warnings" check.
>
> > I did run CI before checkin, and it passed:
> > https://cirrus-ci.com/build/5382423490330624
>
> Weird, why did it not report with the same level of urgency?
> But anyway, thanks for fixing.

Maybe I misunderstood this exchange but ...

Currently Windows warnings don't make any CI tasks fail ie turn red,
which is why Jeff's run is all green in his personal github repo.
Unlike gcc and clang, and MinGW cross-build warnings which cause the
special "CompilerWarnings" CI task to fail (red). That task is
running on a Linux system so it can't use MSVC. The idea of keeping
it separate from the "main" Linux, FreeBSD, macOS tasks (which use
gcc, clang, clang respectively) was that it's nicer to try to run the
actual tests even if there is a pesky warning, so having it in a
separate task gets you that info without blocking other progress, and
it also tries with and without assertions (a category of warning
hazard, eg unused variables when assertions are off).

But I did teach cfbot to do some extra digging through the logs,
looking for various interesting patterns[1], including non-error
warnings, and if it finds anything interesting it shows a little
clickable ⚠ symbol on the front page.

If there is something like -Werror on MSVC we could turn that on for
the main Windows test, but that might also be a bit annoying. Perhaps
there is another way: we could have it compile and test everything,
allowing warnings, but also then grep the build log afterwards in a
new step that fails if any warnings were there? Then Jeff would have
got a failure in his personal CI run. Or something like that.

[1] https://github.com/macdice/cfbot/blob/master/cfbot_work_queue.py