[Bug c++/41892] New: generated code increments past enum boundary

[Bug c++/41892] New: generated code increments past enum boundary

[caolanm at redhat dot com]

i.e. the attached code is fine on x86_64 but crashes on s390x with
g++ -O2 -fpic demo.cxx
it does not crash with 
g++ -O2 demo.cxx
or
g++ -fpic demo.cxx

expected output is...
0 3 0 3 0 3 1 3 2 3 1 3 0 3 1 3 1 3 2 3 2 3 0 3 2 3 1 3 2 3 0 3 0 3 0 3 1 3 2 3
1 3 0 3 1 3 1 3 2 3 2 3 0 3 2 3 1 3 2 3 0 3 0 3 1 3 1 3 2 3 2 3
actual output is...
0 3 0 3 0 3 1 3 2 3 0 3 3 3 0 3 4 3 0 3 5 3 0 3 6 3 0 3 7 3 0 3 8 ... crash


-- 
           Summary: generated code increments past enum boundary
           Product: gcc
           Version: 4.4.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: caolanm at redhat dot com
  GCC host triplet: s390x-ibm-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41892

[Bug c++/41892] generated code increments past enum boundary

[caolanm at redhat dot com]

------- Comment #1 from caolanm at redhat dot com  2009-10-31 17:51 -------
Created an attachment (id=18943)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18943&action=view)
demo code

s390x-redhat-linux-g++ -O2 -fpic demo.cxx


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41892

[Bug c++/41892] generated code increments past enum boundary

[caolanm at redhat dot com]

------- Comment #2 from caolanm at redhat dot com  2009-10-31 17:52 -------
$ gcc --version
gcc (GCC) 4.4.2 20091019 (Red Hat 4.4.2-5)


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41892

[Bug c++/41892] generated code increments past enum boundary

[rguenth at gcc dot gnu dot org]

------- Comment #3 from rguenth at gcc dot gnu dot org  2009-11-01 10:26 -------
You should report this to RedHat, RedHat ships modified GCCs.  Alternatively
reproduce this with a FSF release.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41892

[Bug target/41892] generated code increments past enum boundary

[jakub at gcc dot gnu dot org]

------- Comment #4 from jakub at gcc dot gnu dot org  2009-11-06 10:22 -------
The only difference on the attached testcase between vanilla 4.4 branch and
4.4-RH is unwind info, in particular epilogue unwind info.  The testcase never
throws, so I don't think it matters.
With -fno-exceptions -fno-asynchronous-unwind-tables the generated assembly is
identical (in both cases from x86_64-linux -> s390x-linux cross).
I'm now grabbing a real s390x box to see whether the testcase actually fails at
runtime...


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41892

[Bug target/41892] generated code increments past enum boundary

[jakub at gcc dot gnu dot org]

------- Comment #5 from jakub at gcc dot gnu dot org  2009-11-06 12:35 -------
Anyway, I can't reproduce the runtime failure, neither with HEAD of 4.4 branch,
or 4.4-RH branch, nor 4.4.2-5.el6, nor 4.4.2-7.el6.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41892

[Bug target/41892] generated code increments past enum boundary

[caolanm at redhat dot com]

------- Comment #6 from caolanm at redhat dot com  2009-11-06 16:26 -------
Hmm, I can reproduce it 100% on hercules, but checking on a real s390x and
despite matches the hercules settings to that of the real box it works fine on
reality and blows up on the emulator. 

Lets blame hercules then until there's some evidence to think otherwise. Odd
though.


-- 

caolanm at redhat dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |WORKSFORME


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41892