On Thu, Jan 14, 2021 at 10:51 AM Tom Lane <tgl@sss.pgh.pa.us> wrote:
>
> Michael Paquier <michael@paquier.xyz> writes:
> > On Thu, Jan 14, 2021 at 01:17:57AM +0000, Hou, Zhijie wrote:
> >>>> Thanks. I am thinking to backpatch this even though there is no
> >>>> problem reported from any production system. What do you think?
>
> > text_to_cstring() indeed allocates a new string, so the extra
> > allocation is useless. FWIW, I don't see much point in poking at
> > the stable branches here.
>
> Yeah, unless there's some reason to think that this creates a
> meaningful memory leak, I wouldn't bother back-patching.
>
The only case where it might impact as per the report of Zhijie Hou is
where the user is subscribed to the publication that has a lot of
tables like Create Publication ... For All Tables. Even though for
such a case the memory consumed could be high but all the memory is
allocated in the Portal context and will be released at the statement
end. I was not sure if that could create a meaningful leak to any user
so to be on the safer side proposed to backpatch it. However, if
others don't think we need to backpatch this then I am fine doing it
just for HEAD.
--
With Regards,
Amit Kapila.
