g77 of egcs-1.0.2/1.0.3a returns exit code 1 for argument type mismatch

[Craig Burley]

>Now, what I want to know is that such a treatment for argument type mismatch is intended by g77 developers or not.

Please read up on the -fno-globals and -Wno-globals options in the
g77 docs, as that should answer your questions.  (The diagnostics
related to these options should have pointers to specific chunks
of text in the documentation, as well.)

>> [At least one (rather reliable) proprietary compiler I know of behaves
>> essentially the same as g77 in this context.]
>
>I remember that VAX or CRAY fortran (?) can do run-time argument type check by putting information of argument data type on stack.      

g77 doesn't do that, but it'd be a nice feature, since static checking
(and certainly the limited kind of static checking g77 does) cannot
catch all cases.

>Anyway, compiling in fsplit'ted form is safer (to deceive the compiler :-))...

Yes, if you're a "perfect" programmer who never writes bad code.  :)

However, note that fsplit'ed source, as with compiling with -fno-globals,
basically disables some useful optimizations (mainly having to do
with inlining), and that's likely to remain true for some time.

        tq vm, (burley)