Re: [cc65] testsuite reloaded (die ultimative lobhudelei :))

Hi!

On Mon, Jul 27, 2009 at 05:07:21PM +0200, Groepaz wrote:
> oookay :) some recent ramblings on c.s.c. made me look at my compiler
> testsuite again, i have updated it a bit and i ran it on the latest snapshot
> in various configurations (c89/c99/cc65 standard that is). i also added a few
> more tests that check certain things that were reported as bugs in older
> versions. and guess what? 2.12.9 gloriously passes *every single test*,
> including some of the weird ones that exposed some bugs in earlier versions.
> (and for giggles: including one test that gcc manages to screw up, depending
> on optimization settings). thats awesome, really. many thanks and much
> respect to Uz, and everyone else involved in bringing the compiler to this
> state.

I hadn't seen the "discussion" on c.s.c before. It's ... interesting :-)

Anyway, thanks for the praise. Other people seem to think better abount cc65
than I do. I've been very lazy in the last few years.

> i will put up a webpage some time later with the detailed reports, i'd like to
> add even more tests - and atleast one that breaks something :) - before
> though, since right now the results are a bit "boring" =P (btw, if someone
> knows any regression test collections other than what sdcc and lcc comes
> with - i have included those already - point me to them please)

Here is something that cc65 isn't able to compile, despite the fact that it's
legal C:

        static int i[3][4] = { 1, 2, };

I'm probably able to find more such snippets, but this is the only one I can
remember for now. Actually, I remembered more, but when I tried the others, I
found that they were already fixed :-)

> ... and then said ramblings on usenet mentioned dhrystone.c, which made me
> curious enough to compile it with cc65 too :) and again, the outcome looks
> totally awesome. check this:

Benchmarking is like statistics, you can prove almost anything by applying the
right thing. "Lies, Damn Lies, and Benchmarks" is the technical term, as far
as I remember :-) Anyway, nice to see. Especially when considering the
shortcomings of the current cc65 implementation.

Regards


        Uz


-- 
Ullrich von Bassewitz                                  uz@musoftware.de
----------------------------------------------------------------------
To unsubscribe from the list send mail to majordomo@musoftware.de with
the string "unsubscribe cc65" in the body(!) of the mail.