Re: Discussion about why GNU/Linux system upgrades cause old programs to break

"Dr. Arne Babenhauserheide" <[email protected]> writes:

> If the situation would have been that once all programs were following
> the spec and then an isolated change was made that only enabled
> breaking the spec in ways that are prone to errors, then I’d agree:
> that shouldn’t be done.
>
> But that’s not what happened: before c99 programs were following older
> specs and c99 wasn’t a singular change, so that point is moot.

I'm not sure what you mean by this.

The C99 mode was not made default until long after.  One would have to
specifically opt into it until the mid-2010s.

When you did, you'd immediately get hit with diagnostics for these
cases, just as warnings rather than errors.

To my understanding, it was already common practice to explicitly opt
into -std=c99/gnu99 and fix the ensuing diagnostics by the time when the
default mode was changed (to C11, skipping C99 entirely).

Indeed, as you say, C99 was not an singular change.

But the pre-C99 mode remained default for quite a long time.

Thus, the deviation from the standard was in an *opt-in* C99 mode at the
time; this is why I find it a bit odd.

But, as I said, it very well could've been justified at the time by some
factor that I'm missing right now.  But I do not think it's still
justified in 2024 (when the default was flipped).
-- 
Arsen Arsenović

-----BEGIN PGP SIGNATURE-----

iQECBAEWCgCqFiEE/uKz0RP8AKMWLWBhUsKUMB6ixJMFAmqYCK4bFIAAAAAABAAO
bWFudTIsMi41KzEuMTIsMiwyXxSAAAAAAC4AKGlzc3Vlci1mcHJAbm90YXRpb25z
Lm9wZW5wZ3AuZmlmdGhob3JzZW1hbi5uZXRGRUUyQjNEMTEzRkMwMEEzMTYyRDYw
NjE1MkMyOTQzMDFFQTJDNDkzEBxhcnNlbkBhYXJzZW4ubWUACgkQUsKUMB6ixJMc
wAD9Ei3VL0J36O4Xn88X9nMbrIjcku0JolMsaQFqdQb1Tc8A/jQXUgU+I73xlsha
/+ZWCMq9KCFMdSdriYF956ONquwI
=u1uA
-----END PGP SIGNATURE-----