Re: [boost] metrowerks codewarrior regression failures appeared

From: Beman Dawes (bdawes_at_[hidden])
Date: 2002-12-20 16:21:40

At 03:21 PM 12/20/2002, Samuel Krempp wrote:

 >Checking today's boost.format regression results, there was new failures
 >for the metrowerks compiler.
 >Was it recently updated ? (now 8.3)

Yes, fairly recently. Also the new sane-testing version of testing.jam may 
be reporting some errors that were not correctly reported in the past.

 >I see boost.format is not the only lib for which the codewarrior
 >regression status changed.
 >
 >At least in my case, codewarrior is to blame for the new failure :
 >
 >#   'getloc' is not a member of class 'std::basic_stringbuf<char,
 >std::char_traits<char>, std::allocator<char>>'
 >
 >while a basic_stringbuf should inherit getloc from its basic_streambuf
 >base.

Be careful. Remember that Metrowerks is the only Win32 compiler that 
diagnoses two-phase lookup errors correctly. And that is exactly the 
symptom (in a template, works on other compilers, but Metrowerks can't find 
the symbol) of two-phase lookup errors.

While Dave and others have explained two-phase lookup to the point I have 
some understanding, and can fix simpler errors, this one is beyond me.

The full error message is given below in case others would like to hazard 
an opinion.

--Beman

### mwcc Compiler:
#      In: ..\boost\format\parsing.hpp
#    From: ..\libs\format\test\format_test2.cpp
# ---------------------------------------------
#     371:        while(i1 < buf.size() && 
isdigit(buf[i1],oss_.rdbuf()->getloc())) ++i1;
#   Error: 
       ^
#   'getloc' is not a member of class 'std::basic_stringbuf<char, 
std::char_traits<char>, std::allocator<char>>'
#    (instantiating: 'boost::basic_format<char, 
std::char_traits<char>>::parse(const std::basic_string<char, 
std::char_traits<char>, std::allocator<char>> &)')