Re: BUG #18354: Aborted transaction aborted during cleanup when temp_file_limit exceeded

At Thu, 22 Feb 2024 14:38:15 +0900 (JST), Kyotaro Horiguchi <horikyota.ntt@gmail.com> wrote in 
> At Thu, 22 Feb 2024 11:49:29 +0800, Tender Wang <tndrwang@gmail.com> wrote in 
> > Changing the behavior of tuplestore can work for this issue,  but I'm not
> > sure if this change will affect other components which depend on
> > BufFile(like nodeMaterial)

In the case of nodeMaterial, tuplestore is simply discarded and
underlying files are removed without flushing by the resource owner
mechanism. On the other hand, in the case of this bug report, the
files in the problem tuplestore are not managed by a resource
owner. As a result, PortalDrop have to explicitly call tuplestore_end
to release resources used by the tuplestore.

>     /*
>      * Delete tuplestore if present.  We should do this even under error
>      * conditions; since the tuplestore would have been using cross-
>      * transaction storage, its temp files need to be explicitly deleted.
>      */
>     if (portal->holdStore)
>     {
>         MemoryContext oldcontext;
> 
>         oldcontext = MemoryContextSwitchTo(portal->holdContext);
>         tuplestore_end(portal->holdStore);

As far as I can see, there are no caller sites to tuplestore_end in
the abort path or any error handling path.

regards.

-- 
Kyotaro Horiguchi
NTT Open Source Software Center