This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Patch for c/13801
On Tue, 24 Aug 2004, Zack Weinberg wrote:
> Then we have an impasse, and the logical thing to do is get someone
> else to express an opinion, ideally someone with a good solid
> understanding of the expected semantics of DECLs and their types.
Indeed. (Or more than one person; anyone who can contribute useful
understanding to the issue.)
> If I may attempt to state precisely the point of disagreement: You
> consider the "correct type" of a DECL to be the type specified for
> that DECL by the semantics of the language, at the current point in
> the source program. I consider the correct type to be the one which
> conveys all of the type information available to the compiler about
> that DECL, at the current point in the complete translation, whether
> or not that is the type that the language specifies. Your choice is
> convenient for the front end, mine convenient for the language-
> independent compiler.
This is a fair statement of the disagreement.
A higher-level description would be that we are being perfectionist in
different and conflicting ways (implementing fine points of standards
conformance in the front end versus avoiding certain categories of
possible problems in the middle end).
While I believe we should reduce the front-end / middle-end overlap,
moving towards parsing all translation units (with proper types for front
end) and emitting diagnostics, then gimplifying iff correct (with proper
types for middle end), then passing off GIMPLE to the rest of the
compiler, and this would reduce the scope for any problems associated with
differing expectations of types, I do not believe this should hold up this
patch.
--
Joseph S. Myers http://www.srcf.ucam.org/~jsm28/gcc/
jsm@polyomino.org.uk (personal mail)
jsm28@gcc.gnu.org (Bugzilla assignments and CCs)