Horton

Re: casting operand to proper type in BlockIdGetBlockNumber

Andres Freund <andres@anarazel.de>

View thread View on postgr.es
From: Andres Freund <andres@anarazel.de>
To: Tom Lane <tgl@sss.pgh.pa.us>
Cc: Zhihong Yu <zyu@yugabyte.com>, PostgreSQL Developers <pgsql-hackers@lists.postgresql.org>
Date: 2022-03-03T19:46:00Z
Lists: pgsql-hackers 
Hi,

On 2022-03-03 14:00:14 -0500, Tom Lane wrote:
> The attached is enough to get through check-world with
> "-fsanitize=undefined" using RHEL8's clang 12.0.1.

Cool.


> I'm not sure whether to back-patch --- looking through the
> git logs, it seems we've back-patched some fixes like these
> and not others.  Thoughts?

It'd be easier to run a BF animal if we fixed it everywhere.


> In any case, if we're going to take this seriously it seems like we need a
> buildfarm machine or two testing this option.

I was planning to add it to the CI runs, just didn't have energy to fix the
failures yet. But you just did (although I think there might be failure or two
more on new-ish debians).

For the buildfarm, I could enable it on flaviventris? That runs an
experimental gcc, without optimization (whereas serinus runs with
optimization). Which seems reasonable to combine with sanitizers?

For CI I compared the cost of the different sanitizers. It looks like
alignment sanitizer is almost free, undefined is pretty cheap, and address
sanitizer is pretty expensive (but still much cheaper than valgrind).

Greetings,

Andres Freund

PS: Hm, seems mylodon died a while ago... Need to check what's up with that.

Commits

  1. Tighten overflow checks in tidin().

  2. Remove some pointless code in block.h.

  3. Fix bogus casting in BlockIdGetBlockNumber().

  4. Clean up assorted failures under clang's -fsanitize=undefined checks.