This is the mail archive of the
mailing list for the GCC project.
Re: genrecog: warn of unmatchable predicate codes
> Except that we're actively doing the wrong thing, so... no, I
> think that an error is appropriate. The define_predicate thing
> is new enough it won't affect many ports.
As we're in regression-only mode, and bumping error_count would
immediately kill the i386 build, I disagree. I'm certainly willing to
add the error_count bump after someone patches at least i386 and/or
any other primary release platforms, and verifies that the fixed
predicates don't cause regressions.