This is the mail archive of the
mailing list for the GCC project.
Re: [4.1] UCNs in identifiers
- From: Mike Stump <mrs at apple dot com>
- To: Zack Weinberg <zack at codesourcery dot com>
- Cc: gkeating at apple dot com (Geoffrey Keating), gcc-patches at gcc dot gnu dot org
- Date: Fri, 7 Jan 2005 11:06:15 -0800
- Subject: Re: [4.1] UCNs in identifiers
On Thursday, January 6, 2005, at 09:59 PM, Zack Weinberg wrote:
You may not commit this patch, or any other patch along the same
lines. This feature MUST NOT be implemented until the associated bugs
in the C and C++ standards are fixed.
I find this position somewhat drastic. Our history is littered with
bad language specs, incomplete language specs, contradictory language
specs, missing language specs and beyond. I don't find it at all bad
to implement the middle core that is obvious, agreed upon and implied
if not stated by the standard; failing that, what users want.
I don't find it a crime to change an ABI for a new feature that no one
uses, and no one in their right might could portably use.
I guess I am just more pragmatic. The compiler isn't an exercise in
academic purity, nor necessarily the reflection of some perfection of a
language spec, but rather, what a user wants. Either, they tell us
what they want, or we decide what they want.
I appreciate that you may not want to do the deciding, what's your
right, but, also, I don't see that progress should be blocked, just
because some silly language spec. Rewrite the language specs if you
want, and submit that to the committee if you want.
We can do up a proposed solution that resolves all the known issues,
deliver that to users, get feedback on it, propose it as the standard.
I'd rather the standard be based upon something that someone has
thought through, implemented, resolved all the issues that users bring
up and that users use...