On Sun, 15 Jun 2025, Jon Turney wrote:

> On 14/06/2025 06:14, Jeremy Drake wrote:
> > On Fri, 13 Jun 2025, Jeremy Drake via Cygwin-developers wrote:
> >
> > > I got permission denied trying to push to a new branch.
>
> You have been granted additional permissions which should allow you to push
> now.

Thanks.  Don't know now if I'll use it right away... ;)  I could add a
regression test for cygwin_create_path /dev/sda

> > > I'm working on posix_spawn tests on github, since that's where the actions
> > > would run, and also locally and on Linux.
> > >
> > > https://github.com/jeremyd2019/cygwin-stc/compare/main...refs/heads/posix-spawn
> >
> > Oh, also let me know if you think this sort of test belongs somewhere else
> > (maybe winsup/testsuite? though I don't know how it would work out there)
>
> I don't know.
>
> I'm not a huge fan of the clever tricks which winsup/testsuite does to run
> against the just-built cygwin, but they do work.
>
> (Ofc, when they break, we usually end up silently running the tests against
> the installed cygwin instead, which we could probably do with a guard against)
>
> There's some pain relying on anything else being available (currently we jump
> through some hoops to provide busybox 'sh' and 'sleep' which are used by some
> of the LTP tests...)
>
>
> The advantage of stc setup is that we can build with headers from, and link
> against, the cygwin we're testing against; and install anything else from the
> distro.
>
>
> So I guess pick whichever seems most appropriate to you :)

I think the winapi/testsuite actually makes more sense, these are not
regression tests or simple test cases, but an attempt to pretty
exhaustively exercise an API to *avoid* accidentally introducing
regressions.  It will require me to retool the tests I wrote so far,
splitting into a bunch of small test .c files and invoking /proc/self/exe
with an argument that tells the child that it's the child so it does
whatever simple thing I was invoking env, readlink, or grep to do in the
tests I wrote so far.

Reply via email to