🚀 go-pugleaf

In article <4jcknm$a13@solutions.solon.com>, news@Thinkage.On.CA says...

> Actually, not all compilers follow the above.  Ours have several levels
> of diagnostics.
> 
> Fatal_errors:  The code is illegal, and so badly screwed up that
>        the compiler abandons further source analysis.
> Errors: The compiler detected something illegal and will not continue
>         other passes.  No usable object deck (loader input) will
>         be produced.  The compiler does try to analyse for other
>         errors.
> Warning: The code is illegal, but the compiler thinks it can generate
>         code to do what the programmer wanted.  (The duplicate
>         qualifier probably gets this.)  A usable object deck is
>         produced.

Hmm...an interesting way of going at things.  Given the wording of the 
standard, I'm not sure you can say the implementation isn't conforming because 
of this, but I rather doubt it's exactly what was intended.  Does it behave 
this way regardless of whether ISO conformance is requested?

> There are also some other diagnostics for legal but questionable
> constructs.  (E.g. variables declared but not used) These get
> "notes" and "unusual" type headers.

Ah, interesting.  The one I'm accustomed to simply classifies warnings by 
number - the "level" of warning is the first digit of the warning number.  
Level 1 warnings are nearly always errors while level 4 warnings are only 
slightly more likely to indicate real errors that those from lint. (Things 
like comparing signed to unsigned types, etc.)
    Later,
    Jerry.