Re: Bug in C++ tree inlining or in PPC backend?

[Mark Mitchell <mark at codesourcery dot com>]

>>>>> "Franz" == Franz Sirl <Franz.Sirl-kernel@lauterbach.com> writes:

    Franz> 1. a string 'object' #1 is created in stackslot X 2. more
    Franz> string objects are created, which eventually reuse
    Franz> stackslot X 3. the objects get used 4. object #1 gets
    Franz> deleted/destroyed. If stackslot X got used in 2., that
    Franz> causes the corruption.

I believe you.  That sounds totally plausible.

    Franz> Sure. So I'll continue to wade thru code I don't understand
    Franz> :-). I'll be away til Sunday though.

Thanks.  Best would be if you could send me preprocessed source, the
.rtl file (generated with -g -da), and an explanation of where exactly
in the RTL the problems are occurring.  (Thus and such stack slot
contains the string, see here it's initialized, look here its
clobberred, and look here its used again.)  

We don't have a powerpc system to play with, which is why I'm going to
need lots of help from you, sadly.

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com