O0 cleanup_cfg vs debuggability

[Olivier Hainque]

Daniel Jacobowitz wrote:
> Absolutely.  Don't conditionalize it on -g; conditionalize it on
> !optimize.

 Thanks for the prompt feedback.

 That's actually what I did, which unfortunately doesn't change much because
 no -g also implies no line number notes, whatever the optimize value.

 A possible alternative is to always return false if !optimize, which seems
 a bit too fierce in first approximation.

 The other way out is to avoid edge forwarding if "some_condition",
 with some_condition resolving to !optimize.

 It's certainly possible to literally express some_condition as !optimize. I
 thought some "mode" based expression would have been preferrable, although
 I have no strong feeling about this.

 Currently, a mode based expression would be a test against CLEANUP_EXPENSIVE,
 which I'm not sure would properly convey the real intent from a code reader's
 perspective.  This is why I mentioned something like CLEANUP_CONSERVATIVE in
 the initial message, which could be useful later on if a similar issue
 happens to show up out of another transformation.

 What do you think ?