Re: [Patch,Fortran] PR40863 [4.5 Regression] Build failure in libgfortran

[Janne Blomqvist <blomqvist dot janne at gmail dot com>]

On Wed, Jul 29, 2009 at 17:46, Tobias Burnus<burnus@net-b.de> wrote:
> Janne Blomqvist wrote:
>> On Mon, Jul 27, 2009 at 09:26, Tobias Burnus<burnus@net-b.de> wrote:
>>
>>> c) My recent patch forgot to update gfortran.map; while doing
>>> ? so, I added some other functions which were missing.
>>>
>> Unless I'm misunderstanding something, shouldn't the symbols below
>> belong to the GFORTRAN_C99_1.1 symbol node rather than 1.0?
>>
>
> Probably yes. I do not know when these were added (before 4.3.0 or in
> 4.4?) but in terms of having a @name they will be new in 4.5 and thus I
> think you are right that for linker resolution they should be in
> GFORTRAN_C99_1.1.
>
> I am not sure when I have time for gfortran - presumably not before
> Monday. Thus: Go ahead if you want to correct it yourself.

Committed r150765 as obvious.

>(While you
> are it: Can you add a comment to libgfortran/fmain.f90 that this file is
> obsolete since GCC 4.5 and there only temporarily for backward
> compatibility.)

Similarly, r150767 committed as obvious.

> PS: We should try harder to remember updating the .map file - we
> seemingly didn't for those C99 functions. Have we any GFORTRAN_1.*
> routine which is missing?

Well if a routine isn't in the .map file, it's not visible outside the
shared library. So for normal functionality this should be a
non-issue. The C99 stuff is special since it's just fallbacks, so
those of us with full-featured C99 libraries don't necessarily notice
it if something's missing.


-- 
Janne Blomqvist