[Bug target/48542] mmix code which calls setjmp/longjmp may not work.

[Bug target/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

Hans-Peter Nilsson <hp at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2011.06.04 14:33:24
                 CC|                            |hp at gcc dot gnu.org
         AssignedTo|unassigned at gcc dot       |hp at gcc dot gnu.org
                   |gnu.org                     |
     Ever Confirmed|0                           |1
      Known to fail|                            |4.3.5

--- Comment #3 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-04 14:33:24 UTC ---
Confirmed.  Sorry I hadn't seen this report.  Your patch looks likely but I'll
take a deeper look.

[Bug target/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

--- Comment #4 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-07 05:46:10 UTC ---
(In reply to comment #3)
> Your patch looks likely

Something along those lines might be sufficient and contained for release
branches, but I think it should be differently at least on trunk: the
initial-value machinery should have handled the register with the
return-address being clobbered. It does partially; note the store of that
register at the function prologue (with an unpatched gcc).

[Bug target/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

--- Comment #5 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-14 07:02:02 UTC ---
(In reply to comment #4)
> on trunk: the
> initial-value machinery should have handled the register with the
> return-address being clobbered. It does partially; note the store of that
> register at the function prologue (with an unpatched gcc).

Yep, it's (brr) reload.  I'll attach a patch momentarily.  I'll have to finish
my analysis and do a mmix regtest and a hosted regtest before posting.

[Bug target/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

Hans-Peter Nilsson <hp at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #23938|0                           |1
        is obsolete|                            |

--- Comment #6 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-14 07:12:04 UTC ---
Created attachment 24516
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=24516
Fix for trunk and maybe branches

Patch against trunk r174961.  Fixes in two places (reload.c:find_equiv_reg and
reload1.c:reload_as_needed) are necessary, because when register contents is
invalidated for the main path (the patch to reload1.c), there's fallback code
that also iterates over code, and which also lacks proper handling of
setjmp-type calls. So, you can see this bug on other targets if you're
"unlucky" enough to write code that need reloads and which seems reusable
across a setjmp-call.
This just naturally happens more often for mmix (like always), given that (set
rJ (mem ...)) is emitted after every call and always needs a reload through a
general register, and is always reloaded from the same mem.
(Tentative partial analysis.)

[Bug rtl-optimization/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

--- Comment #7 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-17 02:18:27 UTC ---
Author: hp
Date: Fri Jun 17 02:18:24 2011
New Revision: 175132

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=175132
Log:
    PR rtl-optimization/48542
    * reload.c (find_equiv_reg): Stop looking when finding a
    setjmp-type call.
    * reload1.c (reload_as_needed): Invalidate all reload
    registers when crossing a setjmp-type call.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/reload.c
    trunk/gcc/reload1.c

[Bug rtl-optimization/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

--- Comment #8 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-17 02:20:14 UTC ---
Author: hp
Date: Fri Jun 17 02:20:11 2011
New Revision: 175133

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=175133
Log:
    PR rtl-optimization/48542
    * gcc.dg/torture/pr48542.c: New test.

Added:
    trunk/gcc/testsuite/gcc.dg/torture/pr48542.c
Modified:
    trunk/gcc/testsuite/ChangeLog

[Bug rtl-optimization/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

Hans-Peter Nilsson <hp at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.7.0

--- Comment #9 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-17 02:22:01 UTC ---
Four more branches to go.  Testing native...

[Bug rtl-optimization/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

--- Comment #10 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-17 11:13:42 UTC ---
Author: hp
Date: Fri Jun 17 11:13:38 2011
New Revision: 175139

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=175139
Log:
    PR rtl-optimization/48542
    * reload.c (find_equiv_reg): Stop looking when finding a
    setjmp-type call.
    * reload1.c (reload_as_needed): Invalidate all reload
    registers when crossing a setjmp-type call.

Modified:
    branches/gcc-4_6-branch/gcc/ChangeLog
    branches/gcc-4_6-branch/gcc/reload.c
    branches/gcc-4_6-branch/gcc/reload1.c

[Bug rtl-optimization/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

--- Comment #11 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-17 11:15:39 UTC ---
Author: hp
Date: Fri Jun 17 11:15:35 2011
New Revision: 175140

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=175140
Log:
    PR rtl-optimization/48542
    * gcc.dg/torture/pr48542.c: New test.

Added:
    branches/gcc-4_6-branch/gcc/testsuite/gcc.dg/torture/pr48542.c
Modified:
    branches/gcc-4_6-branch/gcc/testsuite/ChangeLog

[Bug rtl-optimization/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

--- Comment #12 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-17 11:17:37 UTC ---
Author: hp
Date: Fri Jun 17 11:17:34 2011
New Revision: 175141

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=175141
Log:
    PR rtl-optimization/48542
    * reload.c (find_equiv_reg): Stop looking when finding a
    setjmp-type call.
    * reload1.c (reload_as_needed): Invalidate all reload
    registers when crossing a setjmp-type call.

Modified:
    branches/gcc-4_5-branch/gcc/ChangeLog
    branches/gcc-4_5-branch/gcc/reload.c
    branches/gcc-4_5-branch/gcc/reload1.c

[Bug rtl-optimization/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

--- Comment #13 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-17 11:18:39 UTC ---
Author: hp
Date: Fri Jun 17 11:18:37 2011
New Revision: 175142

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=175142
Log:
    PR rtl-optimization/48542
    * gcc.dg/torture/pr48542.c: New test.

Added:
    branches/gcc-4_5-branch/gcc/testsuite/gcc.dg/torture/pr48542.c
Modified:
    branches/gcc-4_5-branch/gcc/testsuite/ChangeLog

[Bug rtl-optimization/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

--- Comment #14 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-17 11:20:05 UTC ---
Author: hp
Date: Fri Jun 17 11:20:01 2011
New Revision: 175144

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=175144
Log:
    PR rtl-optimization/48542
    * reload.c (find_equiv_reg): Stop looking when finding a
    setjmp-type call.
    * reload1.c (reload_as_needed): Invalidate all reload
    registers when crossing a setjmp-type call.

Modified:
    branches/gcc-4_4-branch/gcc/ChangeLog
    branches/gcc-4_4-branch/gcc/reload.c
    branches/gcc-4_4-branch/gcc/reload1.c

[Bug rtl-optimization/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

--- Comment #15 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-17 11:20:41 UTC ---
Author: hp
Date: Fri Jun 17 11:20:38 2011
New Revision: 175145

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=175145
Log:
    PR rtl-optimization/48542
    * gcc.dg/torture/pr48542.c: New test.

Added:
    branches/gcc-4_4-branch/gcc/testsuite/gcc.dg/torture/pr48542.c
Modified:
    branches/gcc-4_4-branch/gcc/testsuite/ChangeLog

[Bug rtl-optimization/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

--- Comment #16 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-17 11:24:57 UTC ---
Author: hp
Date: Fri Jun 17 11:24:55 2011
New Revision: 175146

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=175146
Log:
    PR rtl-optimization/48542
    * reload.c (find_equiv_reg): Stop looking when finding a
    setjmp-type call.
    * reload1.c (reload_as_needed): Invalidate all reload
    registers when crossing a setjmp-type call.

Modified:
    branches/gcc-4_3-branch/gcc/ChangeLog
    branches/gcc-4_3-branch/gcc/reload.c
    branches/gcc-4_3-branch/gcc/reload1.c

[Bug rtl-optimization/48542] mmix code which calls setjmp/longjmp may not work.

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

--- Comment #17 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-17 11:25:56 UTC ---
Author: hp
Date: Fri Jun 17 11:25:53 2011
New Revision: 175147

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=175147
Log:
    PR rtl-optimization/48542
    * gcc.dg/torture/pr48542.c: New test.

Added:
    branches/gcc-4_3-branch/gcc/testsuite/gcc.dg/torture/pr48542.c
Modified:
    branches/gcc-4_3-branch/gcc/testsuite/ChangeLog

[Bug rtl-optimization/48542] unchanged variables in code which calls setjmp may be clobbered (including the return-address)

[hp at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

Hans-Peter Nilsson <hp at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED
   Target Milestone|4.7.0                       |4.3.6
            Summary|mmix code which calls       |unchanged variables in code
                   |setjmp/longjmp may not      |which calls setjmp may be
                   |work.                       |clobbered (including the
                   |                            |return-address)

--- Comment #18 from Hans-Peter Nilsson <hp at gcc dot gnu.org> 2011-06-17 11:38:05 UTC ---
Fixed on all open branches. Plus title change to reflect the general nature of
the bug. Actually s/variables/registers/ but let's keep it user-friendly when
we can.

[Bug rtl-optimization/48542] unchanged variables in code which calls setjmp may be clobbered (including the return-address)

[ebotcazou at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48542

Eric Botcazou <ebotcazou at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |martin at netbsd dot org

--- Comment #19 from Eric Botcazou <ebotcazou at gcc dot gnu.org> 2012-06-17 09:18:29 UTC ---
*** Bug 53219 has been marked as a duplicate of this bug. ***