On Mon, 7 Jul 2025, Corinna Vinschen wrote: > On Jul 7 12:16, Jeremy Drake via Cygwin-patches wrote: > > On Mon, 7 Jul 2025, Corinna Vinschen wrote: > > > All good points. We should actually see what the Austin Group comes up > > > with and then we can reconsider. In the meantime we stick to your current > > > implementation. Would you mind to push it on top of main into a new > > > topic branch, i.e., something like > > > > > > git checkout -b topic/posix_spawn main > > > > > > and push it? If you're not aware of this, the "topic/" prefix is > > > required to allow force pushing to the branch. It's some kind of > > > safety net from the gerrit macros activated for a couple of projects > > > on sware. > > > > Done. > > https://www.cygwin.com/cgit/newlib-cygwin/log/?h=topic%2Fposix_spawn > > > > This also includes the patch I recently sent, because I had done half of > > that while adding pgroup support. > > Looks good. However, shouldn't the hunk adding InterlockedCompareExchange > setting the pgid go into its own patch? That looks more like a bugfix > to me...
I don't think it's a bugfix - previously, this was where the pgid was initialized and it was done unconditionally. Now that I want to set the pgid in child_info_spawn::worker, this needs to not overwrite that already-set pgid. (This does not fix the issue where I see a pgid of 0 in a spawned process sometimes instead of what it should have inherited from the parent, which I assume is a race between the child running and the parent finishing up this initialization).
