Re: [PATCH] Fix socket handle inheritance on Windows

On 10/21/26 08:22, Harrison Booth wrote:
> Bryan,
>
> I've attached your public patch with no changes. Also attached is a
> test-only regression supplement.
>
> On PostgreSQL master 6b9f40d49ae33eb0b4a7c77666fe4971ef7ba087, the
> exact ordered series applied with git am. The focused pg_ctl test
> passed all nine assertions on hosted Windows x64 and ARM64. On the
> unpatched baseline, only the two socket-lifetime assertions failed on
> both hosts. The supplement forces a fresh initdb so the test does not
> copy the temporary install cluster.
>
> Hosted run: https://github.com/Harrison-Booth/postgresql-windows-ci/actions/runs/39594799011
>
> This was a focused test, not a full regression run. It does not
> implement the shared SOCK_CLOEXEC flag or accept4-style API discussed
> in the thread. Could you review whether this test-only supplement
> belongs with your patch as posted?
>
> Best,
>
> Harrison

Harrison,

Thanks for kicking the tires on this. I had this and several other
windows patches on hold due to another issue that has now evaporated. I
will be posting some new patches today or tomorrow. I have made the
test a bit different, but let me post it and then we can discuss.
Really appreciate you taking the time and effort on this!

--
Bryan Green
EDB: https://www.enterprisedb.com