Re: using scratchpads to enhance RTL-level if-conversion: the new patch is almost ready to be prepared for merging to trunk, but not 100% ready yet

[Jeff Law <law@redhat.com>]

On 09/25/2015 01:18 PM, Segher Boessenkool wrote:
> On Fri, Sep 25, 2015 at 10:09:23AM -0600, Jeff Law wrote:
>>>> So what that means is the presence or absence of debug information is
>>>> causing a difference in
>>>> the code you generate.  That is (of course) bad and indicates a bug
>>> of some kind in your code.
>>>
>>>> I haven't put your code under a debugger or anything like that, but
>>>> this does jump out:
>>>
>>>> +        rtx_insn* temp_src_insn = BB_HEAD (then_bb);
>>>> +        if (temp_src_insn && NOTE_INSN_BASIC_BLOCK_P (temp_src_insn))
>>>> +          temp_src_insn = NEXT_INSN (temp_src_insn); /* skip over a
>>>> start-of-BB note */
>>>
>>>> What if BB_HEAD (then_bb) is a DEBUG_INSN with -g enabled, but is a
>>>> NOTE_INSN_BASIC_BLOCK when -g is not enabled.  That could cause
>>>> this kind of failure.
>
> I haven't looked at the full code, but NEXT_INSN is even more suspicious
> (you also need to skip the debug insns).
Yes, you're absolutely right. :-)

jeff