Re: [fortran-dev, patch] Floating-point parser speed-up, once again

[Daniel Kraft <d at domob dot eu>]

Daniel Kraft wrote:
> in light of the new fortran-dev branch and Janne's commit there, I 
> updated my floating-point parser speed-up patch and merged Janne's patch 
> in.
>
> I've yet to do a new regression-testing run, but ok for fortran-dev and 
> 4.5 if successful?

Jerry did give the ok for fortran-dev on IRC, committed with a fix for 
the regression mentioned in this thread to fortran-dev as revision 145259.

> BTW, there's one XXX comment in where I'd like to hear other opinions. 
> If neither blank-zero nor blank-null is selected (that is, 
> blank-unspecified) and there's a blank following the mantissa, at the 
> moment we simply regard the parse as done.  However, if a blank appears 
> during parsing the exponent, we keep going and check that really only 
> blanks follow (otherwise, an error is issued).   This was how the code 
> behaved before (if I understood it correct at least) and how my patch 
> still behaves.  I'm no expert on I/O, but this seems a little 
> inconsistent to me; so what to do?

Kept this as discussed with Jerry, will be resolved on merging back to 
trunk.

BTW, I created gcc/testsuite/ChangeLog.dev for fortran-dev changelog on 
the test-suite.

Cheers,
Daniel

--
Done:  Arc-Bar-Cav-Ran-Rog-Sam-Tou-Val-Wiz
To go: Hea-Kni-Mon-Pri