On Wed, Feb 18, 2009 at 7:46 PM, Mark Evenson <evenson@...> wrote:
> Ville Voutilainen wrote:
>> On Wed, Feb 11, 2009 at 9:16 AM, Mark Evenson <evenson@...> wrote:
>>> Not all the changes I would propose for 0.13 are in the build system:
>>> Right now [TRACE is once again broken on trunk] that should be fixed
>>> before release. And I think there is a rather deep bug somewhere within
>>> ABCL that breaks SLIME CVS HEAD that I would love to resolve. But
>>> abcl-0.12.0 is broken in this way as well (abcl-0.0.11 is the oldest
>>> version that still works).
>> Mark, are there any other things that need to be fixed before release,
>> and do you
>> need help with any of them?
> I still have problem with SLIME HEAD, due to some sort of deadlock
> between the Emacs and ABCL threads about whether a READ should have
> completed or not, but I don't expect anyone else to get deep into SLIME
> to try to debug. I have a workaround, and will try to expand this into
> a more general fix, but I don't think it makes sense to hold up 0.13 on
> this bug.
> Other than that, clearing up the TRACE facility would be wonderful. As
> it stands in the release branch, the TRACE form can be compiled, but
> attempting to trace forms that invoke the compiler (qv. (TRACE FORMAT))
> will eventually fail with unpleasant results.
Well, this has been fixed and backported:
(format t "lisp")
(format t "lisp")
0: (FORMAT T "lisp")
0: FORMAT returned NIL
Next to that, it's now possible to compile TRACE forms. And it's
possible to trace into the compiler.
> Given the bombshell that
> Erik recently discovered in the compiler (as to incorrect coverage of
> exceptions) I'm surprised that anything works at all, so this is by no
> means essential.
Well, anything that doesn't contain UNWIND-PROTECT, CATCH or TAGBODY
should work. Other than that, the most common cases for these also
work. So, it's by no means impossible to create working compiled
systems, but there are definitely a number of cases where this is not
possible with our current compiler.
> Other than that, contributions to 'doc' are welcome for points that you
> think could use wider explanation.
> And when I publish a candidate 'build-abcl.lisp', try it on as many
> Lisps as you can, ok?
Any news or progress you can report on this item? I mean, if you're
having specific problems we can help try figure out solutions or
>> Regarding the license text modification that I did on trunk, do you
>> want to merge
>> it to the release branch or should I do it? (I deleted LICENSE, put GPLv2 text
>> in COPYING and appended Classpath exception as the 13th clause)
> Please merge your version of licensing to the release branch as you see
> fit. At a minimum I will certainly be satisfied by what you commit, as
> you have been appropriately vocal in this error. We all agree it's
> GPLv2 with an exception, right?
Thanks for your efforts!