Re: aggressive fixincludes

[Mark Mitchell <mark at codesourcery dot com>]

>>>>> "korbb" == korbb  <korbb@sourceware.cygnus.com> writes:

    korbb> Sure.  Tell me how to identify the fact that "FlexLexer.h"
    korbb> is a C++ file.  Were it, "FlexLexer.H" or "FlexLexer.h++"
    korbb> or just about anything except "*.h", then it would be
    korbb> ignored.  At the same time, I do not think it crucial that
    korbb> the comments remain.  After all, they _are_ comments and
    korbb> /usr/include/FlexLexer.h still remains in /usr/include for
    korbb> visual reference.  Please, propose a solution.

One thing I am concerned about is that fixincluded files to not get
updated when packages get updated.  So, to prevent skew, we should
keep the number of fixincluded files to a minimum, even if the fixes
do not in any way break the file.  Also, in C++, you can step into
code in header files; it's bad if the header is missing its comments.

As for a solution, finding an `extern "C++"' is a good hint it's not C
file.  (`extern "C"' might appear in a C header that can be included
from C++, but `extern "C++"' is much less likely.)  That would cover
FlexLexer.h.

And anything in /usr/include/g++* (g++,g++-2,g++-3) should be excluded
from this test, as those are C++ files.

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com