[llvm-commits] mips commit regarding RA register problem

[Bill Wendling]

On Sep 17, 2012, at 5:05 PM, "Kotler, Reed" <rkotler at mips.com> wrote:

> I have the whole rest of the mips 16 port sitting in our local repository.
> 
> I can't check it in all at once because it's too big.
> 
> So I need to check in all the pieces and then I can do more  work. This is what I'm trying to do right now.
> 
> What part of this check in is bothering you?
> 
> This code will work forever. It's correct.
> 
> What you complained about is the test for a 0 stack size, which  cannot happen right now because
> I am forcing it to have a non zero stack size.
> 
> I don't want to delete that code because it will become allowable once again when I do the optimization which makes it possible to not save RA at all.
> 
> We CANNOT revert this patch right now and continue our work.
> 
> The code we are checking in works. Why is it a horrible hack? It is correct. it will work until the Sun runs out of fuel and stops heating the planet.
> 
Excellent! Fine. Please keep that patch in there and remove any comments that say otherwise.

> You want us to delete two lines of code that test that stack size is 0?
> 
> What problem can this cause?
> 
Dead code is dead code is dead code. Remove it immediately. Also place the addition by the arbitrary number 16 into the initialization of the variable.

> There are NO bugs caused by this checkin.
> 
Great!

> The horrible bug is what happens without this checkin.

You say it's bug free. Then it's bug free. I expect you to never change this in the future.

-bw