[dmd-internals] dmd commit, revision 663

[Rainer Schuetze]

Brad Roberts wrote:
> On Tue, 7 Sep 2010, Rainer Schuetze wrote:
>
>   
>> Brad Roberts wrote:
>>     
>>> My plan at this point is to replace the makefile+do_test.sh with a c++ or d
>>> executable.  I'm tempted to use D, but that introduces a bootstrapping
>>> concern
>>> as well as a reliability concern.
>>>
>>>   
>>>       
>> I'd also like to see the tests being run from a D program (hey, rdmd is
>> supposed to be used like a script language, so it should be up to the task).
>> The ddmd-project has started building the test suite with such a tool
>> recently, but I think it does not have make functionality yet.
>>     
>
> That's several votes for having a D based driver.  I'm still concerned 
> about the same two issues I already raised:
>
> 1) bootstrapping D for new platforms -- the more that has to work before 
> the tests can be easily executed the more pain involved.
> 2) if the D compiler is broken then there's a chance the test suite could show false success.
>
> Both aren't terribly problematic, just give me pause.
>   
The test suite might not help a lot in bootstrapping: the "compilable" 
tests mostly check semantic analysis (which is independent of the 
platform), not if the object file contains sensible code. The "runnable" 
tests require a working druntime/phobos library, so there must already 
be some more or less working compiler available. But having the point 
where it becomes usable as early as possible is still desirable.

> I'm really shocked at how bad msys and cygwin are compared to linux.  Are 
> windows process startup times really that bad?
>   
I guess the bash-startup-time is bad on windows. The runnable tests take 
about 1min25 when compiled and run from a batch file (no output 
verifications).