- asdf-devel - mailman3.common-lisp.net

[asdf-devel] 1.675 bug
by szergling 15 Apr '10

15 Apr '10

Hi list, I'm having trouble hot-upgrading asdf. My old setup was not checked out of a repository, but does have a file modified date of 2007-06-18. I also use asdf-binary-locations version "0.3.8". To begin with, I have created the following config file (~ is at c:/home) ~/.config/common-lisp/source-registry.conf (:source-registry (:directory "e:/src/ltk-0.90/") (:tree "c:/home/code/lisp/libs/") (:exclude "_darcs") :ignore-inherited-configuration) I then load-file asdf.lisp from Slime. My clisp-2.47 (non-cygwin Windows) is previously already running with asdf, asdf-binary-locations, and many systems loaded (and other random work done). I then tried to load a system that I could previously load just fine, in this case ironclad_0.27. CL-USER> (asdf:load-system :ironclad) ; loading system definition from C:\home\code\lisp\libs\ironclad_0.27\ironclad.asd into #<PACKAGE ASDF0> ;; Loading file C:\home\code\lisp\libs\ironclad_0.27\ironclad.asd ... ; registering #<SYSTEM :IRONCLAD #x26F7286D> as IRONCLAD ; registering #<SYSTEM IRONCLAD-TESTS #x26FE5225> as IRONCLAD-TESTS ;; Loaded file C:\home\code\lisp\libs\ironclad_0.27\ironclad.asd ; No value This gave the following error. TRANSLATE-PATHNAME: replacement pieces (NIL NIL (:DIRECTORY #1="home" #2="code" #3="lisp" #4="libs" #5="ironclad_0.27") "package" "fas" :NEWEST) do not fit into #P"C:\\Documents and Settings\\chewyong\\Local Settings\\Temporary Internet Files\\common-lisp\\**\\*.*" [Condition of type SIMPLE-ERROR] I browsed around in the stack trace (yes clisp stack traces are ugly), and looked deeper into this particular frame: ... 17: <1/1718> #<SYSTEM-FUNCTION TRANSLATE-PATHNAME> [1712] EVAL frame for form (TRANSLATE-PATHNAME (ASDF::DIRECTORIZE-PATHNAME-HOST-DEVICE ASDF::P) ASDF::ABSOLUTE-SOURCE ASDF::DESTINATION) Locals: ASDF::ABSOLUTE-SOURCE = #P"\\**\\*.*" ASDF::ROOT = #P"\\" ASDF::DESTINATION = #P"C:\\Documents and Settings\\chewyong\\Local Settings\\Temporary Internet.. ASDF::SOURCE = T #:LIST-13280 = ((T ..)) ASDF::P = #P"\\home\\code\\lisp\\libs\\ironclad_0.27\\package.fas" #:PATTERN-13281 = (T ..) #:TYPECASE-KEY-13279 = #P"C:\\home\\code\\lisp\\libs\\ironclad_0.27\\package.fas" #:TYPECASE-KEY-13278 = #P"C:\\home\\code\\lisp\\libs\\ironclad_0.27\\package.fas" ASDF::PATH = #P"C:\\home\\code\\lisp\\libs\\ironclad_0.27\\package.fas" ... In particular, I'm guessing that this is related to truenamize (has it taken the wrong branch?) with respect to (line 677 on my version of asdf.lisp) (ignore-errors (return (truename p))) Am I close enough? See the following: CL-USER> (truename (merge-pathnames #P"\\home\\code\\lisp\\libs\\ironclad_0.27\\package.fas")) ==> TRUENAME: File #1=#P"C:\\home\\code\\lisp\\libs\\ironclad_0.27\\package.fas" does not exist [Condition of type SYSTEM::SIMPLE-FILE-ERROR] My package.fas, due to asdf-binary-locations, has been stored previously in ~/code/lisp/libs/ironclad_0.27/clisp-2.47-win32-i386/package.fas I hope this is useful for fixing asdf + a-b-l hot upgrade. I can restart a new Lisp and give full backtraces or other further info, should that be necessary. Cheers.

2 2

[asdf-devel] TRAVERSE being very slow
by Faré 15 Apr '10

15 Apr '10

On a large flat system (automatically generated by XCVB from QRes), TRAVERSE takes a whole lot of time. More like O(n^2) or O(n^3) than either O(n) or O(n log n). I suspect suboptimal algorithms are used somewhere, either in component lookup or in topological sorting. Sigh. The amount of memory consed suggests it's not just lookup time, but hugely inefficient intermediate data structures. (time (length (asdf::traverse (make-instance 'asdf:load-op) (asdf:find-system :qres-ccl-xne))) ==> took 140,642,274 microseconds (140.642270 seconds) to run with 8 available CPU cores. During that period, 140,484,780 microseconds (140.484790 seconds) were spent in user mode 148,010 microseconds (0.148010 seconds) were spent in system mode 996,129 microseconds (0.996129 seconds) was spent in GC. 8,182,948,112 bytes of memory allocated. 2748 [ François-René ÐVB Rideau | Reflection&Cybernethics | http://fare.tunes.org ] Economic illiteracy often leads one to take for wealth creation or cost reduction what is only a forced displacement of activity, with no primary gain, and a lot of secondary costs and negative side-effects. — Faré

2 1

[asdf-devel] Results of an automated study of ASDF practices
by Juan Jose Garcia-Ripoll 13 Apr '10

13 Apr '10

... with my usual trolling http://groups.google.com/group/comp.lang.lisp/msg/4fc15114d6865348 FYI -- Instituto de Física Fundamental, CSIC c/ Serrano, 113b, Madrid 28006 (Spain) http://tream.dreamhosters.com

1 0

[asdf-devel] How can I muffle ASDF2 chatter in one shot?
by Jean-Claude Beaudoin 13 Apr '10

13 Apr '10

Hello all, In ASDF2 (I am still using 1.661) is it possible turn off all "informational" output in one single operation? Looking at the source I see that there is a bunch of :verbose arguments to most (all?) of ASDF main entry points, but each of these have hard coded default values of T. This way one has to pass a ":verbose NIL" pair on each call to ASDF in order to get silence. Couldn't we have a *asdf-verbose* global that would give the default value of those :verbose arguments? Then a single (setq *asdf-verbose* nil) would allow the whole thing to go in "run silent, run deep" mode. Thanks Jean-Claude Beaudoin

2 1

[asdf-devel] ASDF 1.671: feature complete - please test!
by Faré 13 Apr '10

13 Apr '10

Dear all, with ASDF 1.671, I've updated the website and hopefully fixed all the bugs due for ASDF release 2. Could you be so kind as to severely test it on your platform of choice and report bugs? I would also invite implementers and system distributers to test-drive it. If you have a beta branch, I would even invite you to include the new ASDF in your beta branch, assuming it passes your preliminary tests. [ François-René ÐVB Rideau | Reflection&Cybernethics | http://fare.tunes.org ] Reevaluate your ends periodically — if some of them or in contradiction with reality or with each other, abandon or amend them without mercy — and those you keep, pursue without any apology.

4 5

[asdf-devel] new asdf does not like matlisp.asd
by m_mommer＠yahoo.com 11 Apr '10

11 Apr '10

Hi, what follows is, as far as I am concerned, an observation, and not a bug report. Maybe it is an interesting issue to keep in mind. So I upgraded to the new asdf, and observed the following. When I load matlisp with the .asd attached, I get an error that goes like this /usr/bin/ld: cannot open output file /home/mommer/.cache/common-lisp/sbcl-1.0.37-linux-x86-64/home/mommer/dev/matlisp/lib/libmatlisp.so: No such file or directory collect2: ld returned 1 exit status The reason for the ld error is that that ^.*/lib/.* does not exist, and again the reason seems to be this: (slightly edited) -------------- [...] (defsystem :matlisp :components ((:unix-dso "alien code" :pathname "" :dso-name "lib/libmatlisp" ;; barf :components ((:alien-module "BLAS" :pathname "LAPACK/BLAS/SRC/" [...] ---------------- I presume one can defend the new asdf here, and I can fix it for me by removing the "lib/" as it does not matter one iota. But I also would expect this to come up again. Regards, Mario.

5 27

[asdf-devel] Upgrade to ASDF 2 with ASDF-BINARY-LOCATIONS
by Robert Goldman 11 Apr '10

11 Apr '10

If one was to start up a lisp with asdf and asdf-binary-locations loaded, and then installed asdf2? Is this expected to run at all well? Would it be reasonable to trap this case and either refuse to run, or somehow undo asdf-binary-locations? I'd like to be able to do some more ASDF2 testing, but if ASDF2 can't be installed over ASDF 1 + A-B-L, that makes it more difficult... Thanks, r

2 1

[asdf-devel] :logical-hostname
by Juan Jose Garcia-Ripoll 10 Apr '10

10 Apr '10

One of the common problems that ASDF users faced when introducing the new binary location system is finding out where things reside. Another problem we will eventually find is locating the resources associated to an ASDF system which has been replaced by a precompiled version. In general it would be desirable for a system to be able to locate its resources (additional files, icons, bitmaps, etc) using logical pathnames. For that purpose I have coded a small patch that adds a new keyword, :logical-host, to DEFSYSTEM. The keyword defines a logical pathname translation from "**;*.*.*" to "path/to/asdf/or/pathname/argument/**/*.*" This way we may have (defsystem :cl-unicode :logical-host "cl-unicode" ...) and in the sources load the databases such as "cl-unicode:uppercase-table.dat" https://bugs.launchpad.net/asdf/+bug/551222 -- Instituto de Física Fundamental, CSIC c/ Serrano, 113b, Madrid 28006 (Spain) http://tream.dreamhosters.com

4 33

Re: [asdf-devel] new asdf does not like matlisp.asd
by Robert Goldman 09 Apr '10

09 Apr '10

traverse returns a plan, which is a list (ordered sequence) of op - component dotted pairs. Effectively, we can execute the plan by mapping PERFORM over it. This is strictly from memory. Hth. "Faré" <fahree(a)gmail.com> wrote: >Ahem. I admit this is a part of ASDF I am not familiar with. > >What does traverse return? Can you trace every exported defgeneric in >ASDF and attach to a launchpad bug traces of what happens in either >ASDF 1 or ASDF 2? > >Sorry I'm not being very helpful. > >[ François-René ÐVB Rideau | Reflection&Cybernethics | http://fare.tunes.org ] > \|/ ____ \|/ > ~@-/ oO \-@~ > /_( \__/ )_\ > \__U_/ > > > > >On 9 April 2010 23:45, Mario S. Mommer <m_mommer(a)yahoo.com> wrote: >> >> Faré <fahree(a)gmail.com> writes: >>> Does it work better with ASDF 1.672? If not, you might have to define >>> some method for operation-done-p or some such. >> >> No, unfortunately it does not work better. >> >> And operation-done-p is returning T's and NIL's with the new asdf as >> well as with the one packaged with sbcl, but the latter simply does not >> take that as a reason to recompile the lisp files, while the former >> does, or recompiles them for some other reason. >> >> Regards, >> Mario >> >> >> >> _______________________________________________ >> asdf-devel mailing list >> asdf-devel(a)common-lisp.net >> http://common-lisp.net/cgi-bin/mailman/listinfo/asdf-devel >> > >_______________________________________________ >asdf-devel mailing list >asdf-devel(a)common-lisp.net >http://common-lisp.net/cgi-bin/mailman/listinfo/asdf-devel -- Sent from my Android phone with K-9 Mail. Please excuse my brevity.

1 0

[asdf-devel] pathname translations does not like my f2cl asd
by m_mommer＠yahoo.com 09 Apr '10

09 Apr '10

Hi, so here is another one. I can send the .asd file if you are interested, but for now I guess this snippet is enough: (defmethod perform ((o compile-op) (f f77-file)) (apply #'f2cl:f2cl-compile (cons (make-pathname :name (component-name f) :type "f" :defaults (component-pathname f)) (compiler-options f)))) the error is Couldn't load #P"/home/mommer/.cache/common-lisp/sbcl-1.0.37-linux-...": file does not exist. [Condition of type SB-INT:SIMPLE-FILE-ERROR] which is caused by the fact that I give the f2cl compiler a different output path than that expected by the loader. I do not know how to fix this yet, except for issuing (asdf:disable-output-translation). It is probably a matter of RTFM. But I guess that proves the more general point that enabling the output translation by default will break existing setups when upgrading. Regards, Mario

2 2