Hi, On Thu, Nov 06, 2008 at 04:01:13PM -0500, DJ Delorie wrote: > > This new failure seems to have been caused by r141613: > > +2008-11-05 Martin Jambor <[EMAIL PROTECTED]> > + > + PR middle-end/37861 > + * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Don't turn > + pointer arithmetics into array_ref if the array is accessed > + through an indirect_ref. > + > > The m32c-elf target is notoriously picky about getting pointer math > "right" (pointers are 24 bits, not 16 or 32), anyone fiddling with > pointer/array addressing stuff should probably test it to see if it > breaks anything.
I only inhibited one particular fwprop action if one particular condition is met. I was slightly afraid of some generated code quality regressions but it should not have really broken anything. > Martin, can you see if you can reproduce this, and verify whether your > change broke it or not? Or if it's something in m32c that your change > expects to be configured differently? My knowledge of backend and RTL in general is very limited and I have no clue what m32c-elf architecture is and don't seem to have access to any such system. >From the logs I see that the testsuite probably uses a cross-compiled gcc? How do you configure gcc to get such a beats? If I manage to reproduce it, I can have a look at what differences in the IR I caused and where that breaks but that all depends on how much my limited RTL knowledge allows me to find out... Martin
