[Bug libfortran/27740] New: libgfortran should use versioned symbols

[Bug libfortran/27740] New: libgfortran should use versioned symbols

[jb at gcc dot gnu dot org]

Gfortran is slowly growing up. As part of that process, as of gcc 4.2 we should
IMHO be more serious about ABI compatibility.

I have a patch forthcoming that bumps the so version and introduces symbol
versioning.


-- 
           Summary: libgfortran should use versioned symbols
           Product: gcc
           Version: 4.2.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libfortran
        AssignedTo: jb at gcc dot gnu dot org
        ReportedBy: jb at gcc dot gnu dot org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[rguenth at gcc dot gnu dot org]

------- Comment #1 from rguenth at gcc dot gnu dot org  2006-05-23 09:37 -------
Nice.  Confirmed.


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2006-05-23 09:37:55
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[jb at gcc dot gnu dot org]

------- Comment #2 from jb at gcc dot gnu dot org  2006-05-23 15:59 -------
Patch here: http://gcc.gnu.org/ml/gcc-patches/2006-05/msg01186.html


-- 

jb at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
   Last reconfirmed|2006-05-23 09:37:55         |2006-05-23 15:59:45
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[tobias dot burnus at physik dot fu-berlin dot de]

------- Comment #3 from tobias dot burnus at physik dot fu-berlin dot de  2006-08-16 17:44 -------
Isn't this fixed as Steve bumped 4.2/TRUNK to libgfortran.so.2 ?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[jb at gcc dot gnu dot org]

------- Comment #4 from jb at gcc dot gnu dot org  2006-08-16 19:09 -------
(In reply to comment #3)
> Isn't this fixed as Steve bumped 4.2/TRUNK to libgfortran.so.2 ?
> 

Umm, no. Symbol versioning is still not implemented. As you can see in the
thread starting from the patch submission, the patch is waiting for
ISO_C_BINDING.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[fxcoudert at gcc dot gnu dot org]

------- Comment #6 from fxcoudert at gcc dot gnu dot org  2006-10-29 14:20 -------
Created an attachment (id=12507)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=12507&action=view)
Patch implementing namespace separation, to go with symbol versioning


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[fxcoudert at gcc dot gnu dot org]

------- Comment #5 from fxcoudert at gcc dot gnu dot org  2006-10-29 14:19 -------
Hi Janne,

As we've discussed on IRC, the inclusion of the ISO_C_BINDING patch is still
uncertain. For other work (implementation of the IEEE intrinsic modules), I've
felt the need to add an option to gfortran that allows symbol names to begin
with an underscore. With it, we can achieve namespace separation of all
external symbols in libgfortran without waiting for ISO_C_BINDING. So, the
patch I propose here:
  - adds a new -fallow-leading-underscore option, which is not documented so
that I doesn't get used by users
  - use it during compilation of the Fortran routines in libgfortran
(specifics, f2c_specifics and selected_{real,int}_kind) and add the _gfortran_
prefix to all these symbols.

With this patch, every external symbol in libgfortran will have the _gfortran_
prefix (except, of course, the fallback C99 math functions we provide when
they're not available on the system). Does that sound OK to you? Is that enough
so that you can update your versioned symbols patch? (I think, the sooner it
goes in, the better)


-- 

fxcoudert at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |fxcoudert at gcc dot gnu dot
                   |                            |org
   Target Milestone|---                         |4.3.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[jb at gcc dot gnu dot org]

------- Comment #7 from jb at gcc dot gnu dot org  2006-10-29 19:32 -------
Yes, this approach is fine by me. Thanks for doing the work.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[patchapp at dberlin dot org]

------- Comment #8 from patchapp at dberlin dot org  2006-11-04 13:02 -------
Subject: Bug number PR 27740

A patch for this bug has been added to the patch tracker.
The mailing list url for the patch is
http://gcc.gnu.org/ml/gcc-patches/2006-11/msg00194.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[dfranke at gcc dot gnu dot org]

------- Comment #9 from dfranke at gcc dot gnu dot org  2007-04-18 17:06 -------
What happend to this? I can't find the patch in the tracker anymore, but
there's no indication in the ChangeLog(s) that is was applied. The ML
discussion went somewhat off track after a while. Any indication to pursue this
path further?


-- 

dfranke at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |dfranke at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[burnus at gcc dot gnu dot org]

------- Comment #10 from burnus at gcc dot gnu dot org  2007-04-18 18:19 -------
> What happend to this? I can't find the patch in the tracker anymore, but
> there's no indication in the ChangeLog(s) that is was applied.

The last patch was applied, i.e. gfortran uses now _gfortran_ prefixes to for
the libgfortran symbols.

However, what is missing are named symbols. That is: One may change the
library, but one keeps the old version. Using the version information, old
programs get the old version. This is used, e.g., by the glibc. If you use
objdump -t libc.so.6, you find:
sched_getaffinity@GLIBC_2.3.3  <--- old version for old programs
sched_getaffinity@@GLIBC_2.3.4 <--- new interface

This was done in the old patch
http://gcc.gnu.org/ml/gcc-patches/2006-05/msg01186.html
That patch got stuck with ISO_C_BINDINGS. We still don't have them, but we have
_gfortran_ prefixes. Thus one could probably start updating that patch even
though ISO_C_BINDING is not yet in gfortran.


-- 

burnus at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |burnus at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[jb at gcc dot gnu dot org]

------- Comment #11 from jb at gcc dot gnu dot org  2007-04-18 19:05 -------
(In reply to comment #10)
> > What happend to this? I can't find the patch in the tracker anymore, but
> > there's no indication in the ChangeLog(s) that is was applied.
> 
> The last patch was applied, i.e. gfortran uses now _gfortran_ prefixes to for
> the libgfortran symbols.
> 
> However, what is missing are named symbols.

To use the proper term, symbol versioning is not yet implemented. Obviusly,
symbols do have names.

> That is: One may change the
> library, but one keeps the old version. Using the version information, old
> programs get the old version. This is used, e.g., by the glibc. If you use
> objdump -t libc.so.6, you find:
> sched_getaffinity@GLIBC_2.3.3  <--- old version for old programs
> sched_getaffinity@@GLIBC_2.3.4 <--- new interface

Yes. For the longer version see http://gcc.gnu.org/wiki/SymbolVersioning and
links therein.

> This was done in the old patch
> http://gcc.gnu.org/ml/gcc-patches/2006-05/msg01186.html
> That patch got stuck with ISO_C_BINDINGS. We still don't have them, but we have
> _gfortran_ prefixes. Thus one could probably start updating that patch even
> though ISO_C_BINDING is not yet in gfortran.

Actually, we don't need ISO_C_BINDING anymore for this; the original reason was
that the Fortran standard doesn't allow procedure names to begin with "_", so
the idea was to use ISO_C_BINDING to mangle the symbol name (to get the names
to begin with "_gfortran"). However, as explained in #5, we now have
-fallow-leading-underscore (undocumented, for internal use only) option for
exactly this purpose.

The reason why the patch is in limbo (besides me being busy with real life :( )
is that while it worked fine under Linux, it turned out that it doesn't work on
many other platforms since on those platforms the C library don't contain all
the C99 stuff that libgfortran needs and thus libgfortran apparently uses its
own implementations instead. Or something like that. So to make the patch work
one would have to, if it works, make the map file a superset of the needed
symbols for all platforms we support. If that doesn't work, I guess one would
have to generate the map file during build time (ugh). Or just disable symbol
versioning for those #@$!% platforms that can't even get a C99 libc done
correctly.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[jb at gcc dot gnu dot org]

------- Comment #12 from jb at gcc dot gnu dot org  2007-04-18 19:49 -------
> The reason why the patch is in limbo (besides me being busy with real life :( )
> is that while it worked fine under Linux, it turned out that it doesn't work on
> many other platforms since on those platforms the C library don't contain all
> the C99 stuff that libgfortran needs and thus libgfortran apparently uses its
> own implementations instead. Or something like that. So to make the patch work
> one would have to, if it works, make the map file a superset of the needed
> symbols for all platforms we support. 

Actually, it looks like this comparatively simple "superset" strategy could
work, at least according to myself a few months ago:

http://gcc.gnu.org/ml/gcc-patches/2006-11/msg00353.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[patchapp at dberlin dot org]

------- Comment #13 from patchapp at dberlin dot org  2007-04-20 07:45 -------
Subject: Bug number PR 27740

A patch for this bug has been added to the patch tracker.
The mailing list url for the patch is
http://gcc.gnu.org/ml/gcc-patches/2007-04/msg01253.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[jb at gcc dot gnu dot org]

------- Comment #14 from jb at gcc dot gnu dot org  2007-04-20 07:45 -------
Patch here: http://gcc.gnu.org/ml/gcc-patches/2007-04/msg01253.html

Also removed the dependency on PR25709 (ISO_C_BINDING), since we don't depend
on that one anymore for this functionality.


-- 

jb at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|25709                       |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[jb at gcc dot gnu dot org]

------- Comment #15 from jb at gcc dot gnu dot org  2007-04-24 10:09 -------
Subject: Bug 27740

Author: jb
Date: Tue Apr 24 10:08:52 2007
New Revision: 124098

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=124098
Log:
2007-04-24  Janne Blomqvist  <jb@gcc.gnu.org>

        PR fortran/27740
        * configure.ac: New test to determine if symbol versioning is
        supported.
        * Makefile.am: Use result of above test to add appropriate linker
        flags.
        * gfortran.map: New file.
        * configure: Regenerated.
        * Makefile.in: Regenerated.
        * config.h.in: Regenerated.


Added:
    trunk/libgfortran/gfortran.map
Modified:
    trunk/libgfortran/ChangeLog
    trunk/libgfortran/Makefile.am
    trunk/libgfortran/Makefile.in
    trunk/libgfortran/config.h.in
    trunk/libgfortran/configure
    trunk/libgfortran/configure.ac


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740

[Bug libfortran/27740] libgfortran should use versioned symbols

[fxcoudert at gcc dot gnu dot org]

------- Comment #16 from fxcoudert at gcc dot gnu dot org  2007-04-24 10:35 -------
Fixed on mainline.


-- 

fxcoudert at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=27740