Re: [DotGNU]Torture Testing ilrun

[Rhys Weatherley]

Gopal V wrote:

> I started this test mainly to see if I could generate multiple
> extra method sections ... But I got a VerificationException
> (not from ilverify, but only from ilrun) at TryTorture2048.cs
> and TryTorture4096 , but not from TryTorture1024....
> 
> This is activated only if someone puts more than thousand exceptions
> into ONE METHOD , so this is a "<this hole exists>" bug and low priority.

I actually got a core dump rather than a VerificationException,
which I've just fixed.

The verifier/coder will bail out if it runs out of method cache
space during conversion.  This can occur if the CVM or unrolled version
of the method is too big to fit within a single method cache page.

The reason why you are getting a VerificationException is that
"convert.c" doesn't currently distinguish between the out of
space case and the "bad instructions used" case.

Whether this is a bug or a feature, well ... :-)  You have to
draw the line somewhere and say "method too big!".

> And the FooBar() method seems to be verified twice ?.

If the converter runs out of cache space, it will allocate
a new cache page and try again.  If it runs out the second
time, it bails out completely.

Cheers,

Rhys.