[Fwd: (j3.2006) Fortran 2003 compiler features]

[Paul Thomas]

Brooks,
>>
>> I'd like to point out that when only a testcase is changed, then a  
>> full regtest is not needed. You can do:
>>
>> (in directory $(builddir)/gcc) make check-gfortran  
>> RUNTESTFLAGS="dg.exp=gfortran.dg/my_test.f90"
>
> Thanks!  That's very useful to have.
It has been posted to the wiki for more than a year now - 
http://gcc.gnu.org/wiki/TestCaseWriting  :-)


I wonder if we cannot bring this thread to a close now?  Not only has it 
drifted off subject but has entered an area of difficulty about which we 
cannot do very much, at present; namely, how do we do a thorough job of 
developing gfortran when we have no full-time, professional involvement 
and all of the volunteers are working flat-out on other activities?  I 
don't know what the answer is but we are not developing it in this thread.

I suggested to Dominique, off-list, that interventions from him in the 
reviewing stage would be very welcome.  It is true that we are not 
always as thorough as we would like to be and that we depend, to a 
greater or lesser extent, on using trunk as an extended testbed.  If 
Dominique and others like him could flag problems, or even potential 
problems, before submission, these would stand a much better chance of 
being dealt with in a timely fashion.  Speaking personally, I would much 
prefer that somebody asks for a hold for a few days, whilst they look 
into a potential problem, than to have to return to a patch months 
later; largely because the thought process behind the patch/testcase is 
still fresh during the submit/commit cycle.

Best regards

Paul