Fortran with GCC

John Tytgat John.Tytgat at
Wed Sep 7 14:57:02 PDT 2005

Hi David,

Apologies for the late reply.  I've been talking about this with Nick
and in the meanwhile there are a couple of family issues requiring some

In message <2453586.182389-fortran at> you wrote:

> Dear John,
>      Thanks for all your explanations of how to run G77.
>      I have modified our GNU page to bring it up to date:

!GNUF77.!Boot is not needed.
!GNUF77.!Run is not in sync with !GCC.!Run version.  And you better mark
the !Run file as modified version of !GCC.!Run.

BTW, could you change the couple of "RISC-OS" references to "RISC OS" which
is the correct spelling of our beloved OS.

>      The problem I have is that g77 still does not link correctly COMMON data
> set up with a BLOCKDATA procedure in one object file to the same COMMON
> in another object file.
>      Most 'bugs' are possible to work around, as this one is by having all the
> program in one file. The working environment then becomes prohibitvely slow,
> having to recompile hundreds of thousands of lines of code just to change one
> line.

I'm not a Fortran expert, only acquired some knowledge on it 12 years ago.
But I think I understand what the intension is.  Nick isn't sure (anymore ?)
whether this is a compiler or linker problem.  Or an AOF file format problem.
In order eliminate the last option (which would be bad news), I assume you
have access to other Fortran compilers allowing you to generate individual
AOF files of the f.Main99_16 and f.Sub99_16, right ? If so, could you e-mail
me those AOF files so I can have a look what AOF features they are using ?

>      The other 'bugs' in our list on the web page can be got around more
> easily.
>      I attach the trivial example which demonstrates the main problem. First
> set up your G77 (by adding G77/zip to Gcc/zip and clicking on !gcc). Then click
> on 'clr99_16' (it assumes that 'f' is in the same directory as 'clr99_16').
> This compiles the two files in the 'f' directory, links them together and
> executes the resulting image giving the wrong result,
> (NUMBER should be 24 wherever it is printed).
>     The file 'Result' shows what happens when the Obey file 'clr99_16' is
> obeyed here.
>     Otherwise GNU F77 for RISCOS is an extremely useful package and one
> which I shall be happy to demonstrate and recommend to our many users. Also
> I may try to adapt one of our Desktop front-end packages to facilitate
> its use. Desktop is the way to use our environment, not 'Make'.

Pretty strong statement which I'm not sure whether I agree with.  Have you
actually used 'make' running in TaskWindow reporting warnings & errors via
throwback ? At least you have something multitasking while the Obey file
mechanism in !GNUF77 isn't.

Anyway, maybe you can ask support for Fortran in EasyGCC2
<URL:> and/or Sourcery
Maybe those projects are something you're looking for.

John Tytgat, in his comfy chair at home                                 BASS
John.Tytgat at                             ARM powered, RISC OS driven

More information about the gcc mailing list