This is the mail archive of the mailing list for the GCC project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: PR c++/26997 g++ reports misleading error message when the identifier with error occurs earlier on the same line

Manuel López-Ibáñez wrote:

>> I think the idea here is somewhat ugly, but I'm not sure how to do better.
> Ugly? Why? In fact, not avoiding tentative parsing is what makes C++
> parsing diagnostics so horrible and debugging it so much pain and
> effort. 

I have no particular love of tentative parsing.  It's necessary in some
cases (given that C++ requires unbounded lookahead in pathological
cases), but you're absolutely right that if we can use it less, that's a
good thing.  My first priority with the new parser was to make it
correct (in contrast to the old parser, which got confused by various
things), but that's not to say that it can't be tightened up.

What I found a bit conceptually ugly here was that we're depending on
the type possibly casted-to, rather than some syntactic property.  It
means, for example, that we'll handle:

  typedef void T();
  ... (T)x ...

differently than:

  typedef void *T;
  ... (T)x ...

In other words, we're not noticing the form "T()" but rather the fact
that it happens to be a FUNCTION_TYPE.

> I will replace the goto with a flag and resubmit.


Mark Mitchell
(650) 331-3385 x713

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]