[Cryptography] Other obvious issues being ignored?

[Jerry Leichter]

> I'll just repeat that again, gcc will quietly break nearly half of all the
> packages that it compiles....
The C community has always relied on "quality of implementation" as a backup to the standard itself.  In effect, it's adding social pressure to legal restrictions.  This is generally a good thing, as some things that are pretty obvious to all practitioners are hard to get down in the formal language of a spec.

I haven't seen anyone mention that on this thread.  In fact, it seems to be a concept that's disappeared from public discourse.

What the paper you're citing proves is the gcc's "quality of implementation" these days is poor, and that people should choose another compiler.  In fact, many people long ago abandoned gcc for llvm-based compilers.  (I don't know how they are doing these days on the handling of these corner, but sometimes important, cases.)

It's also worth pointing out that just because some behavior is left as undefined by the standard doesn't mean it has to be left undefined by an implementation.  An implementation is free to document exactly what it will do in each of these cases.  I would say an implementation that did so, *and* chose "reasonable" behavior, would get some high marks on "quality of implementation".  (Not that I can think of any compiler that chose this route.)
                                                        -- Jerry