>>> "Paul" == Paul Eggert <[EMAIL PROTECTED]> writes:
> Akim Demaille <[EMAIL PROTECTED]> writes: >> I don't believe we ought to fight against bugs in C++ compilers. >> The code we produce is compliant with a (pretty-)well defined standard. > Yes, we needn't port to ancient C++ circa 1985, That's not C++, that's an ancient dialect. (ISO) C++ is born in 95. > but on the other hand we need to be cognizant of the wide variety > of C++ compilers in practical use today, not all of which strictly > conform to the latest standard. 10 years old!?! Then I'd like to know how old was that compiler. > Suppose, for example, the Bison-generated parser triggered a bug in > the C++ compiler in GCC 3.4 and earlier. If that happened, I'd say we > should alter Bison (assuming it's relatively easy) rather than insist > that everybody upgrade to GCC 4.0 before they or their code-suppliers > can use Bison. Older GCC compilers are still quite widely used; GCC > 2.95.4 and GCC 3.3.5 are the two standard GCC installations on Debian > stable, for example. I definitely agree! But I cannot imagine that, given the state of the regression tests as of today, given the extreme aggressiveness that some C++ tests can have wrt the compiler, that some day, code as simple as this one, can make a compiler fail.
