Re: Patch to fix diagnostic format warnings in cp/ dir

[Gabriel Dos Reis <gdr at integrable-solutions dot net>]

"Kaveh R. Ghazi" <ghazi@caip.rutgers.edu> writes:

|  >  > From: Gabriel Dos Reis <gdr@integrable-solutions.net>
|  >  > 
|  >  > the fix is to supply the location info directly, i.e.: 
|  >  > 
|  >  >   cp_error ("%Hno specifiers", &DECL_SOURCE_LOCATION (extra_arg));
          ^^^^^^^^  
|  >  > 
|  >  > -- Gaby
|  > 
|  > Ok?
|  > 
|  > 2003-06-20  Kaveh R. Ghazi  <ghazi@caip.rutgers.edu>
| 
| Actually following this course yields several extra testsuite errors.
| E.g
| 
| > FAIL: g++.old-deja/g++.jason/crash10.C (test for excess errors)
| > FAIL: g++.old-deja/g++.other/anon4.C  (test for errors, line 13)
| > FAIL: g++.old-deja/g++.other/anon4.C (test for excess errors)
| > FAIL: g++.old-deja/g++.other/enum2.C  (test for errors, line 9)
| > FAIL: g++.old-deja/g++.other/enum2.C (test for excess errors)
| > FAIL: g++.old-deja/g++.pt/friend23.C  (test for errors, line 3)
| > FAIL: g++.old-deja/g++.pt/friend23.C (test for excess errors)
| > FAIL: g++.old-deja/g++.pt/redecl1.C  (test for errors, line 9)
| > FAIL: g++.old-deja/g++.pt/redecl1.C  (test for errors, line 15)
| > FAIL: g++.old-deja/g++.pt/redecl1.C  (test for errors, line 19)
| > FAIL: g++.old-deja/g++.pt/redecl1.C  (test for errors, line 22)
| > FAIL: g++.old-deja/g++.pt/redecl1.C (test for excess errors)
| 
| 
| Some are like this:
|  > g++.jason/crash10.C:8: internal compiler error: unexpected letter
|  >    `H' in locate_error
| 
| So the cp_*_at functions don't seems to accept %H because of the
| locate_error() wrapper.

Yes, that is why I did suggest using cp_error -- not cp_error_at -- in
the situation you descirbed (e.g. trailing arguments)


| I also get errors like this:
| 
|  > g++.other/anon4.C:15: internal compiler error: tree check: expected
|  > class 'd', have 't' (union_type) in fixup_anonymous_aggr, at cp/decl.c:6692
| 
|  > g++.other/enum2.C:14: internal compiler error: tree check: expected
|  > class 'd', have 't' (enumeral_type) in start_enum, at cp/decl.c:13014
| 
| So DECL_SOURCE_LOCATION doesn't work on types.  (Obvious in hindsight?)

1) DECL_SOURCE_LOCATION works only on _DECLs (e.g. variables, functions,
    typedefs, ...).  

2) Every type has a _DECL associated with it: the TYPE_MAIN_DECL.

3) For a overload set, you can get the _DECL for one of the function
   through OVL_FUNCTION 

| Anyway, what should I do?

I would suggest you replace the spurious uses of cp_error_at with
cp_error (as indicated in the previous message) and you use

   &DECL_SOURCE_LOCATION (TYPE_MAIN_DECL (t))

when t is a type and 

   &DECL_SOURCE_LOCATION (OVL_FUNCTION (t))

when t is an overload set.  

|  May I install the original patch?  (That one
| passed the testsuite.)

No.  

Thanks for your effort and patience.

-- Gaby