Re: [Patch, libgfortran] PR 27740 Versioned symbols, bump so version

[Janne Blomqvist <jblomqvi at cc dot hut dot fi>]

On Tue, May 23, 2006 at 06:01:15PM +0200, tobias.schlueter@physik.uni-muenchen.de wrote:
> Janne Blomqvist <jblomqvi@cc.hut.fi> wrote on Tue, 23 May 2006:
> >I also decided to version the f2c_specific and specific_ math
> >functions; is this useful or is it even counterproductive?
> 
> The interface of the former functions is frozen to be f2c compatible,  
> if it changes it's a bug.  The latter should be versioned.

Right.

Looking around, the way others (libstdc++ and libgomp) handle it, is
to have separate version nodes for standard and implementation
specific interfaces (e.g. OMP_2.0 and GOMP_1.0). In the same way we
could have a separate F2C_1.0 version node that would remain unchanged
even if we get new GFORTRAN_* nodes. AFAICS the problem with leaving
the f2c stuff unversioned is that the

local:
    *;

thing in the version script hides all other symbols (which in general
is a good thing).

Attached are a new gfortran.map file and the script used to generate
it.

-- 
Janne Blomqvist

Attachment: gfortran.map
Description: Text document

Attachment: createmap.sh
Description: Bourne shell script

Attachment: pgp00000.pgp
Description: PGP signature