Piotr Stefaniak <postgres@piotr-stefaniak.me> writes:
> On 2016-08-18 20:02, Heikki Linnakangas wrote:
>>> - block = (AllocBlock) malloc(blksize);
>>> + block = (AllocBlock) calloc(1, blksize);
>> I think this goes too far. You're zeroing all palloc'd memory, even if
>> it's going to be passed to palloc0(), and zeroed there. It might even
>> silence legitimate warnings, if there's code somewhere that does
>> palloc(), and accesses some of it before initializing. Plus it's a
>> performance hit.
> I just did a test where I [ this n that ]
I think you are failing to understand Heikki's point. There is no way
we are committing the change depicted above, because (1) it will mask more
bugs than it fixes; (2) it's an enormously expensive way to fix anything;
and (3) it will effectively disable valgrind testing for missed
initializations.
The right thing to do is find out what downstream bit of code is missing
initializing a block of memory it's working on, and fix that localized
oversight.
It'd be useful also to figure out why our existing valgrind testing has
not caught this already. The example you give looks like it surely
ought to be replicated well enough in the standard regression tests.
regards, tom lane
