On Wed, Jul 04, 2001 at 01:22:43PM -0400, Dave MacDonald wrote:
> As my first attempt at contribution to SBCL, I removed the cruft meant
> to support the now-unsupported extension allowing COMPILE-FILE to
> compile more than one file.
> This patch doesn't change existing functionality, but it gave me a
> starting place to learn more about hacking on SBCL.
> There is no test attached, but SBCL continues to bootstrap with this
> change. Please review it -- I'd appreciate any feedback.
OK, thank you. I've merged those changes into sbcl-0.6.12.44.
I didn't feel the need to add tests either. The changes basically just
simplify the implementation of existing functionality, and the
functionality seems reasonably well tested by rebuilding the system,
and by running the existing COMPILE-FILE regression tests.
I made a few changes. Notably:
* The patch deleted the definition of FIND-FILE-INFO, perhaps
because you took me at my word when I wrote
;;; FIXME: This is unnecessarily general cruft now that
;;; we only read a single file in COMPILE-FILE.
Unfortunately the patch left a call to FIND-FILE-INFO in
FIND-ERROR-CONTEXT. It didn't need the full generality of the old
FIND-FILE-INFO, but it did need to be rewritten somehow to work
without FIND-FILE-INFO. So I rewrote it.
* The DECLARE IGNORE in FIND-SOURCE-ROOT seemed wrong, so I
* (IF (PATHNAMEP X) X (PATHNAME X)) seemed equivalent to
just (PATHNAME X), so I simplified it.
* I made some names more explicit/unambiguous. E.g. PNAME became
PATHNAME-DESIGNATOR and PATHNAME, so that
(LET ((PNAME (frob .. PNAME ..))) ..) became
(LET ((PATHNAME (frob .. PATHNAME-DESIGNATOR ..))) ..).
Also, in a related move toward explicitness/unambiguity (not a change
to the patch, but a change which interacts with it) I renamed the
FILE-INFO-valued slot of SOURCE-INFO from SOURCE-INFO-FILE to
William Harold Newman <william.newman@...>
pending stuff from sbcl-devel: none
PGP key fingerprint 85 CE 1C BA 79 8D 51 8C B9 25 FB EE E0 C3 E5 7C