Re: Checking for GCC 4

[Ralf Corsepius]

On Fri, 2006-02-03 at 09:56 +0000, Harlan Stenn wrote:
> > > And I'm not sure which is more fragile, user input or coded checks.

> I don't see a significant difference between something that (apparently)
> works that is based on a version number check, an OS version check, or a
> feature test.
Grumble: The point is: GCC version checks are a broken concept and will
never work, primarily because version strings are subject to change at
any time (GCC vendors are supposed to modify them!), so any GCC version
string pattern match will always fail at some time.

The essential difference to target tuples is: Target tuples are de-facto
standardized (by config.guess and config.sub) and unlike GCC version
strings can't easily be changed (Remember the linux-gnu vs. linux vs.
linux-libc* madness).

Ralf