=46rom the SourceForge.net Site Status Page:
On 2004-08-12 at about 10:00 Pacific the pserver based CVS server
that hosts projects that have a first letter of c,d,g,a,w,x and u
had a hardware malfunction that is currently being worked on.
Projects whose letters match those previously mentioned will not be
able to have their pserver CVS repositories or ViewCVS interface
functional until the issue is resolved.
Last time I checked (about a minute ago), there was no word that this
problem had been fixed, so I've put up a new development snapshot:
http://armedbear.org/j-jar.zip (just j.jar)
This snapshot adds a bit of new functionality in Lisp mode (look on the
Mode menu), thanks to Sam Steingold.
ABCL in this snapshot will almost certainly complain about incorrect
fasl versions, if you have any old .abcl files lying around (say from
Prior to this snapshot, ABCL's response to this situation was to mutter
some obscenities and collapse on the floor (although you did get a
prompt, and you could still do very simple things).
Now, ABCL should print a slightly more helpful "Invalid fasl format"
error message for each stale .abcl it encounters, and then fall back to
loading the corresponding .lisp source file, so you should eventually
arrive at a normal, usable system, albeit a more slowly-running one, at
which point you can (and should) do COMPILE-SYSTEM and restart abcl.
The automatic-fallback-to-corresponding-.lisp-file only happens for
files loaded automatically by the system (by LOAD-SYSTEM-FILE and/or
the autoload mechanism), and not for .abcl files explicitly LOADed by
the user. A "use .lisp file" restart would be nice in that situation,
but that's a job for another morning.
The profiler (more precisely, the time-based version of it) now works
correctly (or at least usefully) with compiled code. So you can do:
(prof:with-profiling () (do-something))
and then, when it finishes:
to see where your electrons have been spending their time. (Yes, it's
This was more or less a side effect of an attempt to get a sane
backtrace with compiled code. That work introduced a performance
penalty at default optimization settings. Code compiled with (optimize
speed) isn't subject to the penalty.
In the last few days, the existence of a working profiler has
facilitated enough performance improvements to more than offset the
initial penalty introduced by changing things to make the profiler (and
backtrace) work in the first place, so abcl should be faster now,
overall. (I think.)
I've been spending most of my time lately on the compiler, and as time
goes on, abcl learns how to compile more things. There are still some
things it can't compile, and that's OK. There may even be things it
thinks it can compile, but in reality it screws up and generates
incorrect code. That's not OK, but I'm not aware of any examples of
this in the wild.
I'm sure I'm forgetting some other recent changes that should probably
be mentioned here, but if they're important, you'll notice them
Thanks for your support.