Re: [lmi] [lmi-commits] master fc01b15e 4/4: Avoid overflow on integer multiplication

[Greg Chicares]

On 5/13/22 15:00, Vadim Zeitlin wrote:
[...]
>  I didn't have to fully digest the messages of the last few commits yet,

Neither have I; otherwise, they'd be concise.

I changed some code to do certain multiplications more carefully,
and I've been running into more surprises than I had anticipated;
some were caused by integer overflow.

> but I'm pretty sure that this means that an old task in my TODO list titled
> "Run lmi under UBSAN" is actually of much higher priority than I thought.
> UBSAN would detect all integer overflows (as well as quite a few other
> things) and so I'd strongly recommend running full lmi system test with a
> build using it.

I don't have time to learn how to use a new tool: I'm too busy removing UB.

>  Should I try at least running the unit tests with UBSAN to see if there
> are any low hanging bugs to fix already?

Sounds like a great idea.

> Or will you do it with the full
> test to kill of them with one stone?

The full system test depends on proprietary product files that
cannot be shared, so you can't do that. But if you demonstrate
how to do it for lmi's unit tests, which are all free, then I
can probably adapt your technique, at far less cost than trying
to figure it out all on my own.