On 19.08.2020 21:50, Pavel Stehule wrote:
Hi
Hi hackers,
More than month ago I have sent bug report to pgsql-bugs:
https://www.postgresql.org/message-id/flat/5d335911-fb25-60cd-4aa7-a5bd0954aea0%40postgrespro.ru
with the proposed patch but have not received any response.
I wonder if there is some other way to fix this issue and does somebody
working on it.
While the added check itself is trivial (just one line) the total patch
is not so small because I have added walker for
plpgsql statements tree. It is not strictly needed in this case (it is
possible to find some other way to determine that stored procedure
contains transaction control statements), but I hope such walker may be
useful in other cases.
In any case, I will be glad to receive any response,
because this problem was reported by one of our customers and we need to
provide some fix.
It is better to include it in vanilla, rather than in our pgpro-ee fork.
If it is desirable, I can add this patch to commitfest.
I don't like this design. It is not effective to repeat the walker for every execution. Introducing a walker just for this case looks like overengineering.
Personally I am not sure if a walker for plpgsql is a good idea (I thought about it more times, when I wrote plpgsql_check). But anyway - there should be good reason for introducing the walker and clean use case.
If you want to introduce stmt walker, then it should be a separate patch with some benefit on plpgsql environment length.
If you think that plpgsql statement walker is not needed, then I do not insist.
Are you going to commit your version of the patch?