Re: g77 compiler bug

[craig at jcb-sc dot com]

>I believe I've found a compile-time bug in the preprocessor
>portion of the g77 compiler.

Actually, you're not using the preprocessor (cpp) in this case.
I think you mean the "front end portion".  The g77 docs explain
some of this terminology.  (In particular, g77 is *not* a
preprocessor in the sense that f2c *is*.)

>I've included an example of the 
>offending code at the end this email.  I compile the code 
>on a HPUX 10.20 OS, system.  The 'uname -a' command yields:
>	HP-UX raptor B.10.20 E 9000/802 

I've reproduced it!  Will try to fix it sometime soon.

>By changing the 'integer*2  nels' to 'integer*4  nels'  the code 
>compiles normally.  I searched your 
>	http://egcs.cygnus.com/onlinedocs/g77_toc.html
>web page for an explaination or rationale for requiring either dimemsion
>variables or parameter variables to be integer*4 instead of integer*2
>(after all the value is only three, 3, no need to store it in 4 bytes),
>but could not find a reference.

INTEGER*2 and other support isn't exactly complete, but it's reasonable
to fix problems like this, or provide decent diagnostics, rather
than crash!

>Other than this minor issue, I've beenimpressed with your product.  
>This compiler is far better than the cc or C++ compilers that we have 
>been using from the HP vendor.  And it provides portable source code to 
>boot!   

Thanks for the feedback!

        tq vm, (burley)