On 08/23/2011 03:42 PM, Gerald Pfeifer wrote: > On Tue, 23 Aug 2011, Richard Henderson wrote: >> I've been trying for 2 days to replicate this with various >> configurations and none have failed. > > Thanks for investigating, and sorry that this proves a worthy opponent! > > > On my testers, the system compiler used for bootstrap is GCC 4.2. > Could that make a difference?
I wouldn't think so. > Also, comparing stage2-libiberty/fnmatch.o and stage3-libiberty/fnmatch.o > I noticed that the latter is more than three times the size. > > objdump -d does not show any difference. objdump -W, however, shows > lots of additional DWARF debugging information. I am attaching a diff > between the two. > > Does that help? Not really. This is an obvious and expected difference, since stage2 is normally compiled *without* debug information, so that we can verify that -g0 vs -g2 doesn't alter code generation. r~