On 04/29/14 08:40, Richard Biener wrote:
This is step two - do GIMPLE stmt and SSA verification when the IL
is in such state. It needs extra conditionals for IPA pass
contexts which may run on inconsistent IL until fixup_cfg had
a chance to run and adjust the IL for noreturn and nothrow
changes (done only by ipa-pure-const).
Bootstrapped on x86_64-unknown-linux-gnu, hopefully testing
succeeds this time ;) (see the bug in tree-eh.c this catched)
Any comments?
Thanks,
Richard.
2014-04-29 Richard Biener <rguent...@suse.de>
* passes.c (execute_function_todo): Move TODO_verify_stmts
and TODO_verify_ssa under the TODO_verify_il umbrella.
* tree-ssa.h (verify_ssa): Adjust prototype.
* tree-ssa.c (verify_ssa): Add parameter to tell whether
we should verify SSA operands.
* tree-cfg.h (verify_gimple_in_cfg): Adjust prototype.
* tree-cfg.c (verify_gimple_in_cfg): Add parameter to tell
whether we should verify whether not throwing stmts have EH info.
* graphite-scop-detection.c (create_sese_edges): Adjust.
* tree-ssa-loop-manip.c (verify_loop_closed_ssa): Likewise.
* tree-eh.c (lower_try_finally_switch): Do not add the
default case label twice.
I like it. Can't see any reason not to go forward with it.
jeff
