Dear Ayush, Sawada-san,
> I agree with the analysis and the fix. I've reviewed the v3 patch and
> it looks good to me. IIUC this bug leads to not behavioral problems
> but to memory leaks. I think we can push it without tests. I'm going
> to push it (including back branches), barring any objections.
V3 LGTM.
One additional idea that re-sending the schema can be avoided for stream_abort
in parallel mode. But it's something like improvements and should be discussed
separately.
Best regards,
Hayato Kuroda
FUJITSU LIMITED
