Re: [boost] problem with WCHAR_MIN using gcc-4.1.1 on hpux11.11

From: Moshe Matitya (Moshe.Matitya_at_[hidden])
Date: 2006-09-21 11:45:35

Michael Haubenwallner wrote:
> 
> On Thu, 2006-09-21 at 17:18 +0300, Moshe Matitya wrote:
> > Michael Haubenwallner wrote:
> > > 
> > > On HP-UX 11.11, there is a weird definition of WCHAR_MIN:
> > > 
> > > $ grep WCHAR_MIN /usr/include/*.h
> > > /usr/include/wchar.h:#             define WCHAR_MIN  (wchar_t)0
> > > 
> > > When using gcc-4.1.1 on this platform, each time 
> WCHAR_MIN is tested 
> > > by preprocessor-directive against zero, g++ spits:
> > > 
> > > gcc-C++-action
> > > gcc-C++../../../bin/boost/libs/regex/build/libboost_regex.a/gc
> > > c/debug/c_
> > > gcc-C++regex_traits.o
> > > In file included from ../src/c_regex_traits.cpp:29:
> > > <snip>/boost_1_34_0rc/boost/regex/v4/regex_traits_defaults.hpp
> > :87:94: error: missing binary operator before token "0"
> > > 
> > > This appears in
> > >  boost/regex/v4/perl_matcher.hpp: line 53 and
> > >  boost/regex/v4/regex_traits_defaults.hpp: line 87
> > > 
> > > Both of them are hidden for aCC with "#if !defined(__HP_aCC)", so 
> > > maybe they should also be hidden for gcc on hpux ?
> > > 
> > > This problem does not appear with gcc-3.4.5 because of this:
> > > 
> > > gcc-3.4.5 (don't exactly know until which version) did not define 
> > > _INCLUDE__STDC_A1_SOURCE as builtin, which enables that 
> definition 
> > > of WCHAR_MIN in /usr/include/wchar.h
> > > 
> > > Initially I had this with 1.33.0, but it is the same for 
> rc-1.34.0.
> > 
> > 
> > The ISO C++ Standard [3.9.1] and the ISO C Standard [7.18.3] both 
> > specify that WCHAR_MIN must be a "constant expression 
> suitable for use 
> > in #if preprocessing directives".  HP/UX's definition of WCHAR_MIN 
> > does not conform with this requirement, as the preprocessor has no 
> > idea what to make of the expression "(wchar_t)0".
> > 
> > You can fix this as follows:
> > 
> > #include <wchar.h>
> > #undef  WCHAR_MIN   /* (wchar_t)0 */
> > #define WCHAR_MIN   0
> > 
> 
> Er... I do not have this problem within my code, but while 
> building boost itself - where within boost is best place to do so ?
> 
> And should I file a gcc-bug ?

Sorry, I'm not sure where would be the proper place in Boost to insert
such a workaround.  Is there someone on this list who can advise?

Also, if I'm understanding this correctly, then this is an HP/UX CRT
library bug, not a GCC bug.

Moshe