June 2004 - slime-devel - mailman3.common-lisp.net

[slime-devel] Daily ChangeLog diff
by Luke Gorrie 25 Jun '04

25 Jun '04

Index: slime/ChangeLog diff -u slime/ChangeLog:1.426 slime/ChangeLog:1.427 --- slime/ChangeLog:1.426 Wed Jun 23 21:32:17 2004 +++ slime/ChangeLog Thu Jun 24 06:01:04 2004 @@ -1,3 +1,13 @@ +2004-06-24 Christophe Rhodes <csr21(a)cam.ac.uk> + + * slime.el (sldb-format-reference-node): fix for when `what' is a + list. + (sldb-lookup-reference,sldb-reference-properties): support + :ansi-cl :section reference types. + + * hyperspec.el (common-lisp-hyperspec-6.0): generalize to work + with section numbers lower than 10. + 2004-06-24 Brian Downing <bdowning(a)lavos.net> * slime.el (slime-repl-send-input): Fixed a subtle difference in

1 0

[slime-devel] [CMUCL] Problem: Binding socket to port 0
by case 24 Jun '04

24 Jun '04

Hi, I try to install slime (HEAD at Thu Jun 24 12:26:02 MDT 2004) on my SGI box. I am running cmucl 18e (see the log below) and it complains about a binding to port 0. As far as I know, the problem comes from swank.lisp where we read: (defun start-server (port-file &optional (style *communication-style*) dont-close) (setup-server 0 (lambda (port) (announce-server-port port-file port)) style dont-close)) \!/ Notice the (setup-server 0 ... \!/ and: (defun setup-server (port announce-fn style dont-close) (declare (type function announce-fn)) (let* ((socket (create-socket *loopback-interface* port)) (port (local-port socket))) \!/ Notice the 'port' use on the first ligne of the let* which here is 0.\!/ ... and as far as 'create-socket' uses ext:create-inet-listener, i guess that this is the problem. here goes the log. hope this helps ----------------------< LOG >----------------------- CMU Common Lisp 18e, running on batou With core: /local/lib/cmucl/lib/lisp.core Dumped on: Thu, 2003-04-10 03:01:10+02:00 on Neptune See <http://www.cons.org/cmucl/> for support information. Loaded subsystems: Python 1.1, target SGI CLOS 18e (based on PCL September 16 92 PCL (f)) * ; Loading #p"/usr/people/case/.conf/emacs/slime/swank-loader.lisp". ;; Loading #p"/usr/people/case/.slime/fasl/cmu-18e/swank-backend.sgif". ;; Loading #p"/usr/people/case/.slime/fasl/cmu-18e/nregex.sgif". ;; Loading #p"/usr/people/case/.slime/fasl/cmu-18e/swank-source-path-parser.sgif". ;; Loading #p"/usr/people/case/.slime/fasl/cmu-18e/swank-cmucl.sgif". ;; Loading #p"/usr/people/case/.slime/fasl/cmu-18e/swank.sgif". Warning: These Swank interfaces are unimplemented: (ALL-THREADS INTERRUPT-THREAD RECEIVE RESTART-FRAME RETURN-FROM-FRAME SEND SPAWN WHO-MACROEXPANDS WHO-SPECIALIZES) T * Error in function CREATE-INET-LISTENER: Error binding socket to port 0: Cannot assign requested address Restarts: 0: [ABORT] Return to Top-Level. Debug (type H for help) (CREATE-INET-LISTENER 0 :STREAM :HOST 2130706433 ...) Source: Error finding source: Error in function DEBUG::GET-FILE-TOP-LEVEL-FORM: Source file no longer exists: target:code/internet.lisp. 0]

2 2

[slime-devel] restarts
by Andras Simon 24 Jun '04

24 Jun '04

Both in ACL (6.2 EE) and LW (4.3.6 PE), selecting restarts that need some user intervention, like USE-VALUE, make slime hang. CL-USER> foo [hit 3 for USE-VALUE] enter an expression which will evaluate to a new value: 1 and then nothing happens. Andras [Fresh CVS slime, Emacs 21.2.1, Linux]

5 7

[slime-devel] Startup hang on threaded SBCL
by Nikodemus Siivola 24 Jun '04

24 Jun '04

The following seems reproducible here with CVS HEAD Slime and SBCL 0.8.11.20 build with threads, using :spawn communication style: 1. start emacs 2. M-x slime 3. Hangs while doing "Initial handshake..." 4. Interrupt with C-c C-c from inferior -> *slime-repl* appears and everything is fine after exit from the debugger vai abort or toplevel restarts. Cheers, -- Nikodemus "Not as clumsy or random as a C++ or Java. An elegant weapon for a more civilized time."

1 1

[slime-devel] Startup problem with cmucl
by Stefan Reichör 24 Jun '04

24 Jun '04

Hi! I tried to start slime on a freebsd system. So I typed M-x slime and got the following: CMU Common Lisp 18e, running on nanni.riic.uni-linz.ac.at With core: /.amd_mnt/johnson/vol/vol0/localF5/lib/cmucl/lib/lisp.core Dumped on: Thu, 2003-04-03 20:07:59+02:00 on cvs2.cons.org See <http://www.cons.org/cmucl/> for support information. Loaded subsystems: Python 1.1, target Intel x86 CLOS 18e (based on PCL September 16 92 PCL (f)) * ; Loading #p"/.amd_mnt/johnson/vol/vol0/home/reichr/open_cvs/slime/swank-loader.lisp". ;; Loading #p"/.amd_mnt/johnson/vol/vol0/home/reichr/.slime/fasl/cmu-18e/swank-backend.x86f". ;; Loading #p"/.amd_mnt/johnson/vol/vol0/home/reichr/.slime/fasl/cmu-18e/nregex.x86f". ;; Loading #p"/.amd_mnt/johnson/vol/vol0/home/reichr/.slime/fasl/cmu-18e/swank-source-path-parser.x86f". ;; Loading #p"/.amd_mnt/johnson/vol/vol0/home/reichr/.slime/fasl/cmu-18e/swank-cmucl.x86f". ;; Loading #p"/.amd_mnt/johnson/vol/vol0/home/reichr/.slime/fasl/cmu-18e/swank.x86f". Warning: These Swank interfaces are unimplemented: (RESTART-FRAME RETURN-FROM-FRAME WHO-MACROEXPANDS WHO-SPECIALIZES) T * Type-error in KERNEL::OBJECT-NOT-TYPE-ERROR-HANDLER: NIL is not of type HOST-ENTRY Restarts: 0: [ABORT] Return to Top-Level. Debug (type H for help) (SWANK-BACKEND::RESOLVE-HOSTNAME "127.0.0.1") Source: ; File: /home/reichr/site-lisp/slime/swank-cmucl.lisp (HOST-ENTRY-ADDR-LIST HOSTENT) 0] Any ideas? Stefan.

3 4

[slime-devel] Daily ChangeLog diff
by Luke Gorrie 24 Jun '04

24 Jun '04

Index: slime/ChangeLog diff -u slime/ChangeLog:1.424 slime/ChangeLog:1.426 --- slime/ChangeLog:1.424 Tue Jun 22 10:04:17 2004 +++ slime/ChangeLog Wed Jun 23 21:32:17 2004 @@ -1,3 +1,38 @@ +2004-06-24 Brian Downing <bdowning(a)lavos.net> + + * slime.el (slime-repl-send-input): Fixed a subtle difference in + sending input to the Lisp introduced in 1.316. The newline was + not getting sent, resulting in the Lisp constantly asking for more + read data. I believe the code has been adjusted to behave the + same as 1.315 with regard to sending newlines. + Also adjusted the `slime-repl-old-input' text property to end just + before the newline, not just after. This causes a gap between + inputs even if no Lisp output appeared in between, so that putting + point on an old line and hitting RET will only call up that line, + and hitting RET in the middle of the current line will send it and + not bring up a confusing combination of all previous input. + Many thanks to Loyd Fueston for pinpointing the date and exact + patch for when this problem was introduced. + +2004-06-23 Brian Downing <bdowning(a)lavos.net> + + * slime.el: Re-added most of Luke's patches from yesterday. It + has the shortened names, uses markers instead of stored `(point)' + values, and `slime-fuzzy-complete-symbol' is an option for + `slime-complete-symbol-function'. + It still string compares the target buffer instead of using + `(buffer-modified-tick)'. + I left the `C-c M-i' keybinding in, as it allows use of the + regular completion as well. If there's an objection to this it + can be removed. + `window-configuration-change-hook' is used if the variable is + present, and ignored it not. This neatly sidesteps its absence in + XEmacs while not killing the functionality for GNU Emacs. + + * doc/slime.texi: Added a command entry and short description for + `C-c M-I, slime-fuzzy-complete-symbol', and added its existence to + the `slime-complete-symbol-function' documentation. + 2004-06-22 Luke Gorrie <luke(a)bluetail.com> * doc/slime.texi: Noted ABCL support.

1 0

[slime-devel] Fuzzy completion
by Brian Downing 23 Jun '04

23 Jun '04

Hello Slime developers, For the past while, I've been using a Mac OS X program called Quicksilver, which is sort of a multi-purpose quick access tool. It features a fuzzy completion algorithm that I've quite grown to like. For example I can hit Command-Space and type "liswp<return>", and it will match and launch "LISpWorks Personal". I decided I wanted to add something like this (only better :) to Slime, to go alongside the existing completion algorithm. It can't really be a replacement as they are good at different things. The existing one is good for incrementally typing and completing a symbol. The fuzzy one is good for typing a short version of the symbol, running the completer, and then picking the right result, which hopefully scored at or near the top. While I've attached a patch, this is more a request for comments than a request to commit the code. As I'll discuss, the Emacs interface is incredibly lacking, mostly on account of this being my first time using elisp for anything but configuration. It's very ugly. Also, the completion core code in Swank should be cleaned up - it still looks somewhat hacky. Finally, everything needs documentation. The elisp code has some chunks from isearch.el. Since they are both GPL, I figured this is okay. Besides, it will probably change anyway. :) ---------------------------------------------------------------------- The low level core of the algorithm are implemented in Swank, just like the old completer. The interface looks very similar: ; SLIME 2004-06-10 CL-USER> (setq *print-pretty* t) T CL-USER> (swank::completions "m-v-" "CL-USER") (("multiple-value-bind" "multiple-value-call" "multiple-value-list" "multiple-value-prog1" "multiple-value-setq" "multiple-values-limit") "multiple-value-") CL-USER> (swank::fuzzy-completions "mvb" "CL-USER") (("multiple-value-bind" 26.588236 ((0 "m") (9 "v") (15 "b"))) ("most-negative-double-float" 12.416667 ((0 "m") (11 "v") (17 "b"))) ("most-positive-double-float" 12.416667 ((0 "m") (11 "v") (17 "b"))) ("named-variable" 10.833333 ((2 "m") (6 "v") (11 "b"))) ("*compile-verbose*" 10.666667 ((3 "m") (9 "v") (12 "b"))) ("environment-variable" 10.555555 ((7 "m") (12 "v") (17 "b"))) ("remove-symbol-profiler" 3.5 ((2 "m") (4 "v") (10 "b")))) The return value is obviously a bit more complicated. :) The fields are (completion score chunks), where "chunks" are a description of where the completion matched. There's a convenience function in Swank that is not used by the Emacs interface to print this information in a human readable form. I used this to tune the scoring algorithm. This shows a little more clearly how the algorithm matches: CL-USER> (swank::format-fuzzy-completions (swank::fuzzy-completions "mvb" "CL-USER")) Multiple-Value-Bind score 26.59 ((0 m) (9 v) (15 b)) Most-negatiVe-douBle-float score 12.42 ((0 m) (11 v) (17 b)) Most-positiVe-douBle-float score 12.42 ((0 m) (11 v) (17 b)) naMed-VariaBle score 10.83 ((2 m) (6 v) (11 b)) *coMpile-VerBose* score 10.67 ((3 m) (9 v) (12 b)) environMent-VariaBle score 10.56 ((7 m) (12 v) (17 b)) reMoVe-symBol-profiler score 3.50 ((2 m) (4 v) (10 b)) NIL Here are some more examples, which might show why I think something like this can be useful. (There is an optional third argument to swank::fuzzy-completions, limit. This limits the number of results, which is important for this completer since completing, say, "a" will match many, many symbols. This isn't currently used from Emacs however.) CL-USER> (swank::format-fuzzy-completions (swank::fuzzy-completions "stdout" "CL-USER" 5)) *STanDard-OUtpuT* score 34.83 ((1 st) (5 d) (10 ou) (15 t)) moST-negative-DOUble-floaT score 22.48 ((2 st) (14 dou) (25 t)) moST-positive-DOUble-floaT score 22.48 ((2 st) (14 dou) (25 t)) leaST-positive-DOUble-floaT score 22.45 ((3 st) (15 dou) (26 t)) leaST-negative-DOUble-floaT score 22.45 ((3 st) (15 dou) (26 t)) NIL CL-USER> (swank::format-fuzzy-completions (swank::fuzzy-completions "prp" "CL-USER" 5)) *PRint-Pretty* score 23.83 ((1 pr) (7 p)) PRint-Profile-list score 23.62 ((0 pr) (6 p)) *PRint-Pprint-dispatch* score 23.48 ((1 pr) (7 p)) *step-PRint-Pretty* score 20.59 ((6 pr) (12 p)) *trace-PRint-Pretty* score 20.56 ((7 pr) (13 p)) NIL CL-USER> (swank::format-fuzzy-completions (swank::fuzzy-completions "norm-df" "CL-USER" 5)) least-positive-NORMalized-Double-Float score 32.31 ((15 norm) ...) least-negative-NORMalized-Double-Float score 32.31 ((15 norm) ...) NIL CL-USER> (swank::format-fuzzy-completions (swank::fuzzy-completions "udc" "CL-USER" 1)) Update-instance-for-Different-Class score 26.30 ((0 u) (20 d) (30 c)) NIL CL-USER> (swank::format-fuzzy-completions (swank::fuzzy-completions "urc" "CL-USER" 1)) Update-instance-for-Redefined-Class score 26.30 ((0 u) (20 r) (30 c)) NIL Despite being expensive (it tries all combinations of matches up to a relatively high cutoff per symbol to maximize the score) and not optimized at all, the completion speed doesn't seem bad at all on compiled Lisps on my 500mhz Powerbook. I haven't tried it on interpreted Lisps, but I imagine it wouldn't be too fun. There are some scoring configuration parameters in swank.lisp that are mostly self-explanitory. Adjust to fit your own symbol naming schemes; these seemed to be the five-minute opinion of #lisp. Some way to set these from Emacs would probably be nice in the future. (defparameter *fuzzy-completion-symbol-prefixes* "*+-%&?<") (defparameter *fuzzy-completion-symbol-suffixes* "*+->") (defparameter *fuzzy-completion-word-separators* "-/.") The scoring algorithm is currently completely static. It would probably be a nifty project to implement some machine learning on it. ---------------------------------------------------------------------- Now, the Emacs interface. Well, it kind of sucks. I bound slime-fuzzy-complete-symbol to C-c M-i in the Slime map. The problem is that since the first letters of the short form aren't necessarily the first letters of the matches, the existing Emacs completion system will fail miserably with this algorithm. What I did was, if there was more than one completion, pop up a window with the results (formatted nicely like above, but with a bold face instead of upcasing). Point starts on the highest scoring result, and the result is immediately placed into the Lisp buffer in place of the short form. Then the user can hit 'n' and 'p' to navigate the list of completions. As they do, the replacement is updated. To pick a completion and keep it, hit Return. To abort and revert to the short form, hit almost anything else. It is a recursive edit, like isearch, so doing anything else in Emacs will abort the completion. It's also implemented very poorly. I don't really like this, but I'm not sure what a better option is. If you have the completion buffer come up normally, it will have embedded state from the Lisp source buffer. If the Lisp source buffer changes before a completion is picked, it may modify the wrong position. And again, it can't really work like the (stateless) normal completion. ---------------------------------------------------------------------- So anyway, I'd appreciate it if people could try it out. If anyone has good interface ideas (or even better, would implement them :) I'd like to hear them, since I'm not so happy with what I've got. Thanks, -bcd -- *** Brian Downing <bdowning at lavos dot net>

2 11

[slime-devel] abcl problem and workaround
by Andras Simon 23 Jun '04

23 Jun '04

If you have trouble starting up (a recent copy of) slime with abcl, try setting *swank-debug-p* to nil. The problem is probably a symptom of a bug in abcl's synonym-stream implementation. (There may be other issues involved.) Andras

2 2

[slime-devel] Jumping to source from debugger in Allegro?
by Peter Seibel 23 Jun '04

23 Jun '04

Is this *supposed* to work? I'm using Allegro 6.2 and some recent CVS update of SLIME. But in the debugger if I hit 'v' I get the message: Cannot find source for frame: #<fd -268443954 in "repl-thread" stack group> Pilot error of some sort? -Peter -- Peter Seibel peter(a)javamonkey.com Lisp is the red pill. -- John Fraser, comp.lang.lisp

2 2

[slime-devel] hyperspec-browse-url
by Marco Baringer 23 Jun '04

23 Jun '04

i have a patch to hyperspec.el which allows one to change the viewing of hyperspec entries. this came up on irc a while ago because some people want to use one browser for "regular" links and another for C-c C-h, there are two ways to achieve this, and i don't know which is better: 1) change hyperspec.el to use a hyperspec-browse-url-function variable which is, normally, set to browse-url. 2) buffer locally set browse-url-function. preferences? -- -Marco Ring the bells that still can ring. Forget your perfect offering. There is a crack in everything. That's how the light gets in. -Leonard Cohen

2 1