attached testcase (700kB/.s.bzip2) exposes compile-time hog. i've canceled assembling on P4-2.8GHz after 45 minutes. (gdb) bt #0 0x0805f3ef in resolve_symbol_value () #1 0x0805f7f3 in resolve_symbol_value () #2 0x080636da in relax_segment () #3 0x080620bb in relax_seg () #4 0x080762b9 in bfd_map_over_sections () #5 0x08062bb7 in write_object_file () #6 0x0804bb2e in main () tested on binutils-2.17.50.0.12.
Created attachment 1901 [details] testcase for time-hog.
Current mainline doesn't appear to have a problem on my Athlon box with 1G ram. $ time gas/as-new /tmp/pr4697.s real 0m1.202s user 0m1.100s sys 0m0.080s
Have you tried binutils-2.17.50.0.17?
(In reply to comment #3) > Have you tried binutils-2.17.50.0.17? .17 works fine, but i'm observing new assertion failures in .17 on sparc-sun-solaris2.9. i'll fill it as new PR.