What is the state of release/optimized compilation?

I try periodically to compile my running D projects with optimization switched on and the compiler always breaks with internal errors (cgcod.c ... etc) which are already reported as I well remember.

Without -O the D compiler works smoothly since roughly version 0.80.

For me optimization is not a big issue yet. I would
oly like to see a 'almost bug-free' compiler.

Any news?

Tamás Nagy

MicroWizard schrieb:
> 
> For me optimization is not a big issue yet. I would
> oly like to see a 'almost bug-free' compiler.
> 
> Any news?
> 
> Tamás Nagy

Please take a look at the thread "Code Optimization" under the same group, i.e. digitalmars.D - it might be helpfull and give you some answers.
If you have found a bug, please write some small code that reproduces it and post it on digitalmars.D.bugs

Best,
Tiago

-- 
Tiago Gasiba (M.Sc.) - http://www.gasiba.de
Everything should be made as simple as possible, but not simpler.

Thanks for the very fast answer, but I'm not so new in this forum ;-)

If these errors could be so easily reproduce, I would have sent reports already. But the error code I got was already reported by someone else.

The question is still open.

Tamás

>Please take a look at the thread "Code Optimization" under the same group, i.e. digitalmars.D - it might be helpfull and give you some answers.
>If you have found a bug, please write some small code that reproduces it and post it on digitalmars.D.bugs
>
>Best,
>Tiago

"MicroWizard" <MicroWizard_member@pathlink.com> wrote in message news:djsq82$18cp$1@digitaldaemon.com...
> Thanks for the very fast answer, but I'm not so new in this forum ;-)
>
> If these errors could be so easily reproduce, I would have sent reports
already.
> But the error code I got was already reported by someone else.
>
> The question is still open.

I can't fix a problem for which I don't have a code example that will reproduce it. I think all the reported internal errors that came with reproducible examples have been fixed, if I missed one, please let me know.

On Fri, 28 Oct 2005 09:37:05 -0700, Walter Bright wrote:

> 
> "MicroWizard" <MicroWizard_member@pathlink.com> wrote in message news:djsq82$18cp$1@digitaldaemon.com...
>> Thanks for the very fast answer, but I'm not so new in this forum ;-)
>>
>> If these errors could be so easily reproduce, I would have sent reports
> already.
>> But the error code I got was already reported by someone else.
>>
>> The question is still open.
> 
> I can't fix a problem for which I don't have a code example that will reproduce it. I think all the reported internal errors that came with reproducible examples have been fixed, if I missed one, please let me know.

Walter do you have any idea what you test coverage is for the compiler ? Thats how I find out about code problems which isn't tested.

"Knud Sørensen" <12tkvvb02@sneakemail.com> wrote in message news:pan.2005.10.29.08.02.46.857818@sneakemail.com...
> On Fri, 28 Oct 2005 09:37:05 -0700, Walter Bright wrote:
> > I can't fix a problem for which I don't have a code example that will reproduce it. I think all the reported internal errors that came with reproducible examples have been fixed, if I missed one, please let me
know.
> Walter do you have any idea what you test coverage is for the compiler ? Thats how I find out about code problems which isn't tested.

It wouldn't be hard to write a code coverage analyzer for D, which would tell you which lines of code were executed and which weren't by the test suite. I agree that such a tool would substantially improve the testing.

But that in itself isn't good enough to ensure no bugs.

What I do is add every bug report to the test suite. Over time, this adds up to a killer test suite. The C/C++ one, for example, has 20+ years of bug reports folded into it. Having it gives me pretty good confidence that I can upgrade the compiler without breaking things.

On Sat, 29 Oct 2005 13:10:24 -0700, Walter Bright wrote:

> 
> "Knud Sørensen" <12tkvvb02@sneakemail.com> wrote in message news:pan.2005.10.29.08.02.46.857818@sneakemail.com...
>> On Fri, 28 Oct 2005 09:37:05 -0700, Walter Bright wrote:
>> > I can't fix a problem for which I don't have a code example that will reproduce it. I think all the reported internal errors that came with reproducible examples have been fixed, if I missed one, please let me
> know.
>> Walter do you have any idea what you test coverage is for the compiler ? Thats how I find out about code problems which isn't tested.
> 
> It wouldn't be hard to write a code coverage analyzer for D, which would tell you which lines of code were executed and which weren't by the test suite. I agree that such a tool would substantially improve the testing.

Yes, this is one of the tools missing for D, but I were not talking of test coverage for D code.


> But that in itself isn't good enough to ensure no bugs.
> 
I know, but if some code is not tested it might fail after a update
without your knowing.
Also almost 100% test-covered code forces you to write testable code.


> What I do is add every bug report to the test suite. Over time, this adds up to a killer test suite. The C/C++ one, for example, has 20+ years of bug reports folded into it. Having it gives me pretty good confidence that I can upgrade the compiler without breaking things.

What I were wondering about is if you have run test coverage on the compiler using one of the c/c++ coverage tools.

OK. The bug still exists. (Internal error: ..\ztc\cgcod.c 1489)

I'll dig to the root and will post the result.

Tamas

>I can't fix a problem for which I don't have a code example that will reproduce it. I think all the reported internal errors that came with reproducible examples have been fixed, if I missed one, please let me know.