Jean-Marc Lasgouttes wrote: >>>>>> "Angus" == Angus Leeming <[EMAIL PROTECTED]> >>>>>> writes: > > Angus> I obviously haven't updated the boost part of my LyX 1.3.x tree > Angus> for a while because I just noticed this change being applied to > Angus> my tree today. Why did you add '#include <config.h>' directly > Angus> to boost/config/user.hpp? In the 1.4 tree we have '#define > Angus> BOOST_USER_CONFIG <config.h>' in config.h itself. > > Because in 1.3.x we set some BOOST variables by modifying user.hpp, > whereas this is done in src/config.h in 1.4.x. So the situation is > slightly different. > > The sad part is that this allows to compile on OSX 10.3, but 1.3.7cvs > still crashes at runtime. If you know a tiny little bit about C++ > streams, I'd be open to hints about why this happens. > > See this thread for a backtrace. > http://www.mail-archive.com/lyx-devel AT lists.lyx.org/msg83150.html
Ouch! diff -p -N -r -U 4 13x/boost/boost/format devel/boost/boost/format shows me that the Boost.Format library has been completely rewritten since LyX 1.3 was released. Looking at the backtrace itself, this strikes me: debugbuf::xsputn Why debugbuf? Interestingly, my Linux Alpha machine runs Debian unstable and has g++ 4.0.3 installed. $ grep debugbuf /usr/include/c++/4.0.3/streambuf $ So, is this some Apple addition? Didn't I read something subsequently that upgrading to g++ 4.0.1 cured the problem for Bennett? -- Angus
