On Tue, May 20, 2003 at 03:06:44PM +0100, Ingvar wrote:
> During my recent stay on a small island off Malta, I had a laptop and a recent
> CVS check-out of SBCL with me (and, naturally, an inclination to stay out of
> the sun and instead poke at code).
> The net result is that I now have a compiler where *some* calls to
> give-up-ir1-transform are tagged and will not issue a compiler-note if the tag
> is on a list of "tags to ignore" (at compile-time). It seems as if the code
> is non-broken, I did learn more of the innards of the SBCL compiler and
> I had fun.
> There is also something that is close to a test harness for compilation
> messages (at the moment, it basically checks that everything from the
> third-or-so line to the last-but-one line is the same or different).
> Is there any interest of having diffs propagated back? It seems as if I
> currently have some problems getting at the anon CVS on sourceforge, so I
> can't even give you an estimate of the size of the patches, but I do recall
> them being distributed over a fair number of places, with substantial changes
> to GIVE-UP-IR1-TRANSFORM and IR1-TRANSFORM.
Yes, please send them in. The compiler being too noisy with
optimization notes is a perennial complaint. I've said from time to
time that the solution is nonobvious, and Christophe has sometimes
talked about classifying notes in a way which sounds similar to your
tagging. Whether or not we merge your patches, I'm virtually certain
that we'll be interested to think about them. (And when/if sourceforge
cvs is operational, generating patches and sending them in should only
take you a few minutes.)
William Harold Newman <william.newman@...>
"Keep your noses to the grindstone. The greatest joy is the joy of
duty. Work, work, work." -- Lectroid PA system, _Buckaroo Banzai_
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C B9 25 FB EE E0 C3 E5 7C