This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: Testsuite ad RedHat 7


On Tue, Oct 10, 2000 at 08:11:57PM +0200, pfeifer@dbai.tuwien.ac.at wrote:
> One of the major problems with Red Hat's decision is the fact that this
> makes for three compilers within one year that break binary compatibility
> in one way or the other: GCC 2.95.2, GCC 2.96 (Red Hat) and GCC 3.0.

Perhaps this is a concern for developers, but do you think end users
care??  Even as a developer, I've come to *assume* that C++ releases are
never binary compatible because I loath investigating monsterously
cryptic linker errors.  I always prefer be *sure* about the compiler
version.

This occured to me yesterday: why can't G++ add some magic to each
object file that checks the binary API version?  Something like:

struct gcc_compiled_with_check {
  static void assert_is_compatible(int version);
  gcc_compiled_with_check() { assert_is_compatible(3); }
};
static gcc_compiled_with_check __gcc_compiled_with_check;

I'm not sure about the exact mechanics, but this would nicely eliminate
the ambiguity.  I just hate running code that "almost works" only to
realize 4 hours later that I accidentally mismatched compiler versions.

For what it's worth, $0.02, etc.

-- 
May the best description of competition prevail.
      (via, but not speaking for Deutsche Bank)

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]