[Bug target/29997] New: various targets: GCC fails to encode epilogues in unwind-info

[Bug target/29997] New: various targets: GCC fails to encode epilogues in unwind-info

[amylaar at gcc dot gnu dot org]

See the discussion started with this message:
http://gcc.gnu.org/ml/gcc/2005-10/msg00808.html

According to Jim Wilson in
http://gcc.gnu.org/ml/gcc/2005-10/msg00858.html,
we now have the infrastructure to properly encode prologues / epilogues.
The ia64 port is using this.


-- 
           Summary: various targets: GCC fails to encode epilogues in
                    unwind-info
           Product: gcc
           Version: 4.3.0
            Status: UNCONFIRMED
          Keywords: wrong-code, meta-bug, EH, wrong-debug
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: amylaar at gcc dot gnu dot org
 BugsThisDependsOn: 18749


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29997

[Bug target/29997] various targets: GCC fails to encode epilogues in unwind-info

[pinskia at gcc dot gnu dot org]

------- Comment #1 from pinskia at gcc dot gnu dot org  2006-12-02 08:19 -------
I think this is really a dup of bug 12990.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29997

[Bug target/29997] various targets: GCC fails to encode epilogues in unwind-info

[amylaar at gcc dot gnu dot org]

------- Comment #2 from amylaar at gcc dot gnu dot org  2006-12-04 14:51 -------
(In reply to comment #1)
> I think this is really a dup of bug 12990.
> 

Sort of, but not quite.  12990 does no cover the exception handling aspects,
but it claims that the problem persists target-independently for debug.
But this is not a problem for ia64.  Since we have now the target-independent
infrastructure and one target port that uses it to handle this problem,
this has become a target problem; each target needs to be adjusted separately
to create the proper epilogue information, so it makes sense to track
progress (or lack thereof) for each target individually, while having
one meta-bug to keep track of them all.

Should I mark 12990 as blocking 29997, and 29998 as blocking 12990?

OTOH, when you read the description of 12990, it sounds like the problem
exists only for i[3456]86-*-*, so maybe we should add this to the target
field of 12990 and only make 12990 as blocking 29997.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29997

[Bug target/29997] various targets: GCC fails to encode epilogues in unwind-info

[pinskia at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29997

--- Comment #3 from Andrew Pinski <pinskia at gcc dot gnu.org> 2011-12-22 03:26:11 UTC ---
I think we encode proligue and epilogues now for all targets.

[Bug target/29997] various targets: GCC fails to encode epilogues in unwind-info

[joseph at codesourcery dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29997

--- Comment #4 from joseph at codesourcery dot com <joseph at codesourcery dot com> 2011-12-22 12:53:39 UTC ---
On Thu, 22 Dec 2011, pinskia at gcc dot gnu.org wrote:

> I think we encode proligue and epilogues now for all targets.

It's been done for several targets, but when looking recently I got the 
impression that it hadn't been done yet for SH, for example.

[Bug target/29997] various targets: GCC fails to encode epilogues in unwind-info

[olegendo at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29997

Oleg Endo <olegendo at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |olegendo at gcc dot gnu.org

--- Comment #5 from Oleg Endo <olegendo at gcc dot gnu.org> ---
Any hints on what is missing or needs to be done for SH?

[Bug target/29997] various targets: GCC fails to encode epilogues in unwind-info

[ian at airs dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29997

--- Comment #6 from Ian Lance Taylor <ian at airs dot com> ---
All the insns in sh_expand_epilogue need to be examined to see if they need
REG_CFA notes.  Some of them already have them.  I don't know what more are
needed.

For example, look at the changes made to the i386 backend when rth added the
general support for unwinding through epilogues:
http://gcc.gnu.org/ml/gcc-patches/2009-05/msg00277.html