On Fri, 2006-02-03 at 00:00 -0500, Paul Smith wrote: > %% Ralf Corsepius <rc040203@xxxxxxxxxx> writes: > > rc> On Thu, 2006-02-02 at 16:35 -0500, Paul Smith wrote: > >> %% Harlan Stenn <stenn@xxxxxxxxxxx> writes: > > hs> Sometimes it's Really Hard to check for features/problems. > hs> Really Really Hard. > > >> This is especially true when cross-compiling. > > rc> Simply not true, esp. in case of compilers. > > It IS true. > > rc> Compilers are always running natively on the host. > > Yes, but many features of the compiler cannot be tested just by > compiling code: you actually have to run the code to determine whether > it works one way or the other way. Which compiler or language feature requires you to run code on a target? > Of course there are some obvious things that are syntax related that can > be checked at compile time, but definitely not everything. Examples please. Note: We are talking about compiler (== toolchain) features and packages suffering from the toolchain having changed it's behavior. Compilers never, ever run anything on a target. They are self-contained native programs and do not/are not able to use any piece of information to be retrieved at run-time from the target. Ralf _______________________________________________ Autoconf mailing list Autoconf@xxxxxxx http://lists.gnu.org/mailman/listinfo/autoconf