On Fri, Oct 3, 2025 at 7:31 PM Alexander Korotkov <aekorotkov@gmail.com> wrote:
> On Thu, Sep 11, 2025 at 1:59 AM Thomas Munro <thomas.munro@gmail.com> wrote:
> > FWIW I'm working on a patch set that kills all backends without
> > releasing any locks when the postmaster exists. Then CVs and other
> > latch-based stuff should be safe in this context. Work was
> > interrupted by a vacation but I hope to post something in the nexts
> > couple of days, over on that other thread I started...
>
> How is it going?
I have something, but I ran into a cluster of related subproblems
along the way (namely: how we manage subprocesses for COPY, archiving
etc, which is all a bit of a mess with known problems relating to
interrupts, signals, postmaster exit and [in development] threads,
which I'll write about soon with references to earlier discussions).
I realised that I needed to step back a bit and tackle all aspects of
our process tree and event management in a more architecturally
coherent way, and already had some prototypes for bits and pieces of
that puzzle from earlier attempts, but needed to make them work on
Windows which had some fun subproblems. I think I've mostly figured
it out now and am testing... Unfortunately a planned family vacation
fell in the middle of all that, hence delay. I'm back and actively
working on this now. More very soon.
