On 4/22/19 12:45 AM, Eli Zaretskii wrote:
>
> You have consistently removed the references to -gdwarf-N compiler
> option, presumably on the assumption that GCC nowadays automatically
> uses the highest supported version?

No, I was merely assuming that -gdwarf-4 is good enough to debug Emacs;
-gdwarf-5 is still marked as "experimental" in GCC 8's doc, and I
haven't used -gdwarf-5. If there's a significant advantage to using
-gdwarf-5 then discussion of it should be added to etc/DEBUG. In the
meantime I tried to improve the wording about this, in the attached patch.


> I think this section could greatly benefit from a short explanation
> what "undefined behavior" means, or a couple of examples of it.

It's hard to be brief as -fsanitize=undefined is a bit tricky, but I
took a shot at it by installing the attached.