On Thu, Apr 9, 2009 at 6:40 PM, Freeland Abbott <[email protected]> wrote:
> Isn't that achieved by the auxiliary warnings?  Or do you mean you want them
> all listed in the one error message, or am I misunderstandinging the
> scenario you describe?

Mainly it looks helpful to put it in one place.  I picture the
scenario where someone goes "I though type Foo was serializable,
what's wrong"?  The natural place to look is the error message for Foo
itself.  What do you think?  It seems slightly harder to me to do a
search for "reached via" and then on each one figure out if it was
reached as a subtype of Foo, or reached because it's a field of Foo,
etc.

Additionally, I'm not actually sure the reached via information is
going to be complete.  The same type can be reached via multiple
paths, but only one path is recorded.  If I understand this correctly,
then you might search for "reached via Foo" and not find types that
really were reached via Foo.  You'd have to go do "show hierarchy" in
your IDE, but that's lossy; you won't know exactly what
parameterizations STOB picks for the subtypes.

So, in short, the forward-link messages strike me as better for a
couple of reasons.  That said, it's not a big deal compared to the
change of picking log levels after the dust has settled.  If the
change looks like a pain, or if you prefer the back-link style
messages, then let's haggle about the actual error messages at a later
iteration.  More important is to figure out what to do with the
visited field....

Lex

--~--~---------~--~----~------------~-------~--~----~
http://groups.google.com/group/Google-Web-Toolkit-Contributors
-~----------~----~----~----~------~----~------~--~---

Reply via email to