Hi Sergey,
>Why would the macros from <sys/sysmacros.h> come up after the inclusion of <string>?
I suspect because there is a dependency on <sys/sysmacros.h> from <string>, probably through <cstring> or some other lower level header file.
>I'd think they would be #undef'ed somewhere in the C++ include files that include the <sys/sysmacros.h>.
Why? Each header file is responsible for it's own identifiers; it's not responsible for second guessing all the identifiers that may-or-may-not have been defined or declared by other header files.
>If anywhere at all, they could be accessible from the <c*> include files (if they are covered by the C standard).
System header files are not covered by the C standard.
>I can already see a large list at the top of every sourcefile :) :
I despise the C preprocessor.
I wish that if a programmer wanted to use a preprocessor directive, it'd be explicit, you'd have to do something like this:
#define VALUE 3 int i = VALUE#;
But alas, it's far too late for that to happen.
>And this only if the compiler is g++ 3.x, and #undef'ing an undefined macro isn't polite, so I'd wrap it all in a big ugly #ifdef or a lot of smaller but nevertheless ugly #ifdefs.
As far as I am aware, It's not impolite to #undef an identifier that has not been defined.
>Isn't there another solution?
Write a little program that pulls out every single identifier in your code and makes a paranoia file "foo.paranoia":
#define return return
#define class class
#define minor minor
#define i i
Then do this at the bottom of your source file: #include "foo.paranoia"
That will help you discover any identifier collisions with macros.
BTW: I have seen a #define return macro. That took days to figure out.
HTH, --Eljay
