On 10/21/2010 10:09 AM, Walter Bright wrote:
> 
> 
> Brad Roberts wrote:
>> Walter, this is one of the reasons I groaned when you told me you hadn't
>> actually TESTED the change you made to the new test suite.  Shame shame shame on
>> you.
>> 
> 
> Actually, I did test it, just with my runner, which doesn't run with -O.

Then you didn't test the same change.. since you enabled -O as part of the argument combinations.

Anyway.. it's currently still broken, and that's the important part.  Getting walter to use the public test suite is important, but slightly less so.

So someone needs to work on a fix.  Normally I'd say revert, but behavior changes between -O and not are pretty serious, imho.

On 21 October 2010 20:19, Brad Roberts <braddr at puremagic.com> wrote:
> On 10/21/2010 10:09 AM, Walter Bright wrote:

> So someone needs to work on a fix. ?Normally I'd say revert, but behavior changes between -O and not are pretty serious, imho.

I don't think it's a regression. I haven't seen any evidence that this
ever worked.
The odd thing is that it seems to be Linux only. Bug 4750
"fail_compilation/fail225.d causes dmd to segv" is another Linux-only
fault in the test suite, but it fails in the front-end, so I don't see
how they could be related.
There's nothing that could be reverted, other than the -O.

A similar bug did show up on Windows when I first added those tests to interpret.d. Bug 3264 "-O causes wrong "used before set" error when using enum" Unfortunately that was fixed at the end of the Dark Ages, when there's no use of svn. So it's difficult to work out exactly how it was fixed.