Re: BUG #13594: pg_ctl.exe redirects stderr to Windows Events Log if stderr is redirected to pipe

On Thu, Jan 7, 2016 at 6:42 AM, Alvaro Herrera <alvherre(at)2ndquadrant(dot)com> wrote:
> Michael Paquier wrote:
>> On Mon, Aug 31, 2015 at 11:56 PM, Egon Kocjan wrote:
>
>> Thanks for the additional tests, that's still a win so this patch is a
>> good thing. Relying on isatty is definitely something that we should
>> treat as a bug here and backpatch. On master and REL9_5_STABLE, what I
>> sent previously, with the call of pgwin32_is_service fixed in pg_ctl.c
>> is fine. For back branches, we could simply copy pgwin32_is_service in
>> pg_ctl.c and rely on that. It seems like the safer approach. Thoughts
>> from other hackers on the matter?
>
> I looked at applying this patch, but as it turns out we don't have the
> backpatchable version. I don't think it's a good idea to backpatch the
> move of the file from backend to src/port -- there's precedent for
> copying the entire function (and assorted infrastructure) to pg_ctl
> instead. Would you submit a patch to do that?

Yes (like restricted tokens and pg_upgrade stuff we did), but I was
just waiting that we reach an agreement for the version of master
before moving on with patches for back branches.

> Here's a rebased version of what you submitted last. I tweaked a few
> lines here and there, nothing major. I thought your patch is ready to
> apply and we only need the backpatchable versions.

This rebased patch looks fine for me. I am attaching as well versions
for back-branches for 9.1~9.5. In this case it is just necessary to
duplicate pgwin32_get_dynamic_tokeninfo and pgwin32_is_service to
pg_ctl.c to make things work properly.
Regards,
--
Michael