[Bug middle-end/55035] reload1.c:3766:41: error: âorig_dup[0]â may be used uninitialized in this function (for fr30, microblaze, moxie, rl78)

["law at redhat dot com" <gcc-bugzilla at gcc dot gnu dot org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=55035

Jeffrey A. Law <law at redhat dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |law at redhat dot com
             Blocks|                            |19794

--- Comment #6 from Jeffrey A. Law <law at redhat dot com> ---
Per this discussion:

https://gcc.gnu.org/ml/gcc-patches/2015-07/msg01996.html

It's believed the reduced testcase in c#5 is an invalid reduction.

A cursory review of the code in c#4 indicates that testcase is a valid
reduction.  We ought to be able to determine that the writes inside the two
inner loops do not clobber recog_data.n_dups and thus the first and last loops
iterate over the same space and every read of orig_dup[] in the last loop will
have had a value set in the first loop.

There's almost certainly a hideously complex missed jump threading opportunity
in here.  Conceptually it'd be exposed by duplicating the two inner loops, one
duplicate would be reached when n_dups is zero the other when n_dups is nonzero
after the first loop.   The first duplicate will bypass the last loop the
second duplicate would execute the final loop.  The net result of all that
copying and cfg transformations should in theory make the false positive
warning go away.


Referenced Bugs:

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=19794
[Bug 19794] [meta-bug] Jump threading related bugs