This is the mail archive of the
mailing list for the GCC project.
Re: [patch] for PR 14229
> Looking at the cvs history, I see that the !onlyjump_p test has been
> there every since revision 1.1 of the cfgrtl.c file. It was originally
> in the tidy_fallthru_edges function. Your big 2004-01-29 patch moves
> the test from tidy_fallthru_edges to rtl_tidy_fallthu_edge, and that
> caused try_simplify_condump to fail.
I have moved it from tidy_fallthru_edges in order to make it
ir-independent, so that delete_unreachable_blocks can be reused at
tree level. I have no idea why the test was here; seems that it
was here just always. Handling of the onlyjump was added to
tidy_fallthru_edge (probably) by
2000-04-26 Richard Henderson <firstname.lastname@example.org>
* flow.c (tidy_fallthru_edge): Don't delete the jump when it's
a still-valid conditional jump.
I will assume for now that the test in tidy_fallthru_edges was obsoleted by it
and commit the patch; if it will break something, we will at least learn
what it is (it passes bootstrap & regtesting on i686, ia64, ppc and
hppa, so I assume it won't happen).