gfortran doesn't support visibility.

[Ed Smith-Rowland]

Andrew Pinski wrote:
> Sent from my iPhone
>
> On Apr 5, 2008, at 20:44, Ed Smith-Rowland <3dw4rd@verizon.net> wrote:
>
>> I am forwarding an email I sent about -fvisibility not working for 
>> gfortran.
>>
>> This was marked 35827, a middle end bug.  The gist is that only 
>> C-family languages are deemed to support visibility.  My issue is 
>> that g77 supported visibility and it seems a reasonable notion and a 
>> useful thing to have.  It would keep the symbol table lean and 
>> prevent users from calling routines way down where they shouldn't be.
>
> It did???  -fvisibilty= support did not come in until GCC 4.0 which 
> did not have g77.
It may be that RedHat had patched this is as a vendor extension.  IIRC 
they did have something to do with visibility being added.

But I can't get -fvisibility to work on:
x86_64 Linux
PPC MacOSX
i386

for gfortran 4.3 or 4.1.

I am going to try to find out how C/C++ handles the -fvisibility flag.  
I've never mucked about in this part of the compiler before.  (I mostly 
hack the C++ library ;-) ).

Any pointers would be very helpful.

If I could get the g77 vendor sources that might help a little too.

Ed
>
>>
>>
>> Is there any reason to suppose that gfortran *couldn't* or 
>> *shouldn't* support visibility?  Is there a Fortran language 
>> supported thing that would take the place of this?
>>
>> At least a flag that would mark all methods in a file as having a 
>> specified visibility (i.e. -fvisibility) would help.
>>
>> I know pragmas and __attribute__ and such are not part of Fortran as 
>> they are in C (hence the reclassification as not a Fortran bug?) but 
>> just the per source file flag would help a lot.
>>
>> Here is the original report and the response:
>> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35827
>>
>> Regards,
>>
>> Ed Smith-Rowland
>>
>