[Bug c++/55340] New: Invalid maybe unitialized warning

[Bug c++/55340] New: Invalid maybe unitialized warning

[jwillemsen at remedy dot nl]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55340

             Bug #: 55340
           Summary: Invalid maybe unitialized warning
    Classification: Unclassified
           Product: gcc
           Version: 4.7.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: jwillemsen@remedy.nl


Created attachment 28698
  --> http://gcc.gnu.org/bugzilla/attachment.cgi?id=28698
gzip with both preprocessed files

We have in some of our tests a warning that a variable may be used unitialized,
but we have validated that this is not the case (as far as we could tell). 

We tried to reduce our code to a minimum, but the strange thing is that when we
remove some not related code, the warning is gone. 

Attached 2 preprocessed files, t.cpp t2.cpp, compile using:
g++ --std=c++0x -fno-strict-aliasing -fvisibility=hidden
-fvisibility-inlines-hidden -O3 -ggdb -m64 -pthread -Wall -W -Wpointer-arith
-pipe -D_GNU_SOURCE    -c t.cpp

When compiling t.cpp we get:
In file included from testC.cpp:26:0:                                           
testC.h: In member function ‘void Test::union_union::swap(Test::union_union&)’: 
testC.h:764:3: warning: ‘*((void*)(& intermediate)+8).Test::unionType::disc_’
may be used uninitialized in this function [-Wmaybe-uninitialized]              
testC.h:1325:27: note: ‘*((void*)(& intermediate)+8).Test::unionType::disc_’
was declared here                                                               
When compiling t2.cpp there is no warning. If you check the diff between t.cpp
and t2.cpp you will see that the code that triggers the warning is still there.

We have reduced the code to the point that removal of any type would not
trigger the warning.

[Bug c++/55340] Invalid maybe unitialized warning

[jwillemsen at remedy dot nl]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55340

--- Comment #1 from Johnny Willemsen <jwillemsen at remedy dot nl> 2012-11-15 14:17:38 UTC ---
Lowering optimization from -O3 to -O2 removes the warning

[Bug c++/55340] Invalid maybe unitialized warning

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55340

Paolo Carlini <paolo.carlini at oracle dot com> changed:

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

--- Comment #2 from Paolo Carlini <paolo.carlini at oracle dot com> 2013-02-27 16:03:22 UTC ---
Mainline (would be 4.8.0) is fine and I can't even compile the testcase with
4_6-branch, thus this isn't a regression.

In the future, please do your best to reduce the size of your testcase: in
99.999 % of the cases, 2-3 KB are enough.

[Bug c++/55340] Invalid maybe unitialized warning

[jwillemsen at remedy dot nl]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=55340

--- Comment #3 from Johnny Willemsen <jwillemsen at remedy dot nl> 2013-02-27 19:20:24 UTC ---
Thanks for checking this, will also try 4.8.0 at some point. We have hundreds
of tests using C++11 extensively and this was the only one exposing this
warning. This is mostly generated code and I tried to reduce it further, but
when I reduced it further the warning was gone. I didn't had a clue why it
could be triggered, that lack of knowledge prevented me from reducing the test
case further