This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: C++ PATCH for `throw'
>>>>> "Paul" == Paul Burchard <burchard@pobox.com> writes:
Paul> It looks to my untrained eye like case STMT_EXPR: in
Paul> tsubst_copy() might still be blurring the boundaries between
Paul> tree manipulation and RTL generation phases.
I see why you're fooled; the variable `rtl_expr' is horribly named,
now. It used to be an RTL_EXPR, but now it's just a STMT_EXPR with
the statements instantiated, rather than uninstantiated. I've changed
`rtl_expr' to `stmt_expr' in my local tree, and will check this in
soon.
Paul> (If I understand correctly, that's the boundary you're
Paul> working to clarify, both to enable more-global analysis, and
Paul> also to allow different middle-ends (e.g. doc generators) to
Paul> be plugged in place of RTL generation. Sorry if I'm way off
You're right on the money.
Paul> base here...despite the clarity of ir.texi, I still don't
Paul> understand how this is supposed to happen on the API level.)
That's 'cause I didn't attempt to document that. Way too hairy. :-)
If we can get the representation documented in a way that average
rocket scientists can make sense out of it, that will be major
progress.
--
Mark Mitchell mark@codesourcery.com
CodeSourcery, LLC http://www.codesourcery.com