Andres Freund <and...@anarazel.de> writes:
> On 2016-02-15 14:37:28 +0100, Martin Liška wrote:
>> I've been currently working on support of -sanitize=use-after-scope in the
>> GCC compiler and
>> I decided to use postgresql as my test-case. The sanitation poisons every
>> stack variable at the
>> very beginning of a function, unpoisons a variable at the beginning of scope
>> definition and finally
>> poisons the variable again at the end of scope.
> Generally sounds like a good check.
>> Following patch fixes issues seen by the sanitizer. Hope it's acceptable?
>> With the patch applied, ASAN (with the new sanitization) works fine.
> But I'm not immediately seing why this is necessary? Is this about
> battling a false positive?
I bet a nickel that this is triggered by the goto leading into those
variables' scope ("goto process_inner_tuple" at line 2038 in HEAD).
That probably bypasses the "unpoison" step.
However, doesn't this represent a bug in the sanitizer rather than
anything we should change in Postgres? There is no rule in C that
you can't execute such a goto, especially not if there is no
initialization of those variables.
If you can think of a reasonable refactoring that gets rid of the need
for that goto, I'd be for that, because it's certainly unsightly.
But I don't think it's wrong, and I don't think that the proposed patch
is any improvement from a structured-programming standpoint.
regards, tom lane
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
