On 8/29/15 Aug 29 -3:53 AM, Faré wrote:

I've tried the Allegro 10.0 release candidate, and the script-support bizarrely fails to override the configuration when it passes an argument initialize-source-registry, that gets lost on its way to FLATTEN-SOURCE-REGISTRY, being replaced by NIL:

0[10]: (INITIALIZE-SOURCE-REGISTRY (:SOURCE-REGISTRY :IGNORE-INHERITED-CONFIGURATION)) 1[10]: (COMPUTE-SOURCE-REGISTRY (:SOURCE-REGISTRY :IGNORE-INHERITED-CONFIGURATION)) 2[10]: (ASDF/SOURCE-REGISTRY:FLATTEN-SOURCE-REGISTRY NIL)

I'm a bit too tired to reduce a test case, but look if this is what causes your issue.

This looks like a problem, but it does not look like *my* problem. I get this behavior on Allegro 10's alisp, which is one of the versions of Allegro which *does* pass the tests for me.

I think I have a minimal script for testing this. We should probably start a ticket for this, and I'll put the script there. Then we can report it to Allegro for fixing.

Looks like the PARAMETER local variable in COMPUTE-SOURCE-REGISTRY is somehow being corrupted.

cheers, r

—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org The best place for a striker, that yellow noxious mosquito, is in the concentration camp. — Leon Trotsky, Pravda, February 12, 1920.

On Fri, Aug 28, 2015 at 9:13 AM, Robert Goldman rpgoldman@sift.net wrote:

...

The test-configuration.script and test-logical-pathname.script are failing on the latest Allegro 10 preview on Mac OS X on all the Allegro modern variants, and on the ANSI SMP variants. All tests pass on alisp non-SMP and alisp8 non-SMP.

Test configuration:

Running test-configuration.script with acl-10.0.pre-final.30s-macosx-x64 ; Loading /Users/rpg/lisp/asdf/test/test-configuration.script subdirectories of dir2/: (#P"/Users/rpg/lisp/asdf/build/dir2/dir3/" #P"/Users/rpg/lisp/asdf/build/dir2/dir4/") subdirectories of dir5/: (#P"/Users/rpg/lisp/asdf/build/dir5/dir6/" #P"/Users/rpg/lisp/asdf/build/dir5/dir7/") Source Registry: #<EQUAL hash-table with 22 entries @ #x1000cb1762> TEST ABORTED: the assertion (FIND-SYSTEM :FOO0 NIL) failed.

test-logical-pathname:

Test logical pathnames in source-registry, non-recursive TEST ABORTED: Component :TEST-LOGICAL-PATHNAME not found

Is this the known no-trailing slash problem?

Cheers, r

On 8/29/15 Aug 29 -3:53 AM, Faré wrote:

I've tried the Allegro 10.0 release candidate, and the script-support bizarrely fails to override the configuration when it passes an argument initialize-source-registry, that gets lost on its way to FLATTEN-SOURCE-REGISTRY, being replaced by NIL:

0[10]: (INITIALIZE-SOURCE-REGISTRY (:SOURCE-REGISTRY :IGNORE-INHERITED-CONFIGURATION)) 1[10]: (COMPUTE-SOURCE-REGISTRY (:SOURCE-REGISTRY :IGNORE-INHERITED-CONFIGURATION)) 2[10]: (ASDF/SOURCE-REGISTRY:FLATTEN-SOURCE-REGISTRY NIL)

I'm a bit too tired to reduce a test case, but look if this is what causes your issue.

This looks very weird: Configuring ASDF Tracing INITIALIZE-SOURCE-REGISTRY COMPUTE-SOURCE-REGISTRY FLATTEN-SOURCE-REGISTRY 0[2]: (ASDF/FIND-SYSTEM:INITIALIZE-SOURCE-REGISTRY (:SOURCE-REGISTRY :IGNORE-INHERITED-CONFIGURATION)) 1[2]: (ASDF/SOURCE-REGISTRY:COMPUTE-SOURCE-REGISTRY (:SOURCE-REGISTRY :IGNORE-INHERITED-CONFIGURATION)) ;;; This is done by a format statement in COMPUTE-SOURCE-REGISTRY ;;; Looks like the variable is getting corrupted somehow Before DOLIST, PARAMETER is: NIL 2[2]: (ASDF/SOURCE-REGISTRY:FLATTEN-SOURCE-REGISTRY NIL) 2[2]: returned ((#P"/Users/rpg/common-lisp/" :RECURSE T :EXCLUDE #1=(".bzr" ".cdv" ".git" ".hg" ".pc" ".svn" "CVS" "RCS" "SCCS" "_darcs" "_sgbak" "autom4te.cache" "cover_db" "_build" "debian")) (#P"/Users/rpg/.local/share/common-lisp/systems/" :RECURSE NIL :EXCLUDE NIL) (#P"/Users/rpg/.local/share/common-lisp/source/" :RECURSE T :EXCLUDE #1#) (#P"/usr/local/share/common-lisp/systems/" :RECURSE NIL :EXCLUDE NIL) (#P"/usr/local/share/common-lisp/source/" :RECURSE T :EXCLUDE #1#) (#P"/usr/share/common-lisp/systems/" :RECURSE NIL :EXCLUDE NIL) (#P"/usr/share/common-lisp/source/" :RECURSE T :EXCLUDE #1#)) ;;; format statement in loop In DOLIST, ENTRY is: NIL In DOLIST, ENTRY is: NIL In DOLIST, ENTRY is: NIL In DOLIST, ENTRY is: NIL In DOLIST, ENTRY is: NIL In DOLIST, ENTRY is: NIL In DOLIST, ENTRY is: NIL 1[2]: returned 0[2]: returned

The code for COMPUTE-SOURCE-REGISTRY is this: (defun compute-source-registry (&optional parameter (registry *source-registry*)) (format t "Before DOLIST, PARAMETER is: ~S~%" parameter) (dolist (entry (flatten-source-registry parameter)) (format t "In DOLIST, ENTRY is: ~S~%" parameter) (destructuring-bind (directory &key recurse exclude) entry (let* ((h (make-hash-table :test 'equal))) ; table to detect duplicates (register-asd-directory directory :recurse recurse :exclude exclude :collect #'(lambda (asd) (let* ((name (pathname-name asd)) (name (if (typep asd 'logical-pathname) ;; logical pathnames are upper-case, ;; at least in the CLHS and on SBCL, ;; yet (coerce-name :foo) is lower-case. ;; won't work well with (load-system "Foo") ;; instead of (load-system 'foo) (string-downcase name) name))) (cond ((gethash name registry) ; already shadowed by something else nil) ((gethash name h) ; conflict at current level (when *verbose-out* (warn (compatfmt "~@<In source-registry entry ~A~@[/~*~] ~ found several entries for ~A - picking ~S over ~S~:>") directory recurse name (gethash name h) asd))) (t (setf (gethash name registry) asd) (setf (gethash name h) asd)))))) h))) (values))

it certainly looks like something is going wrong handling optional arguments.

On 8/29/15 Aug 29 -2:03 PM, Kevin Layer wrote:

...

...

it certainly looks like something is going wrong handling optional arguments.

Robert,

If you can get a test case, that would wonderful. I appreciate the help narrowing down this problem.

Kevin

I pushed a topic branch "allegro-10-bug". Look for test/test-allegro10.lisp. You'll have to modify the *asdf-directory* location at the top of the file appropriately (mine is hardcoded in there).

Then I do:

rlwrap /usr/local/acl/acl100pf.64/alisp (load "test-allegro10.lisp")

and you should be able to replicate the issue.

R

Because I'm not an ASDF developer and know little of it, I was hoping for a test case that involved primarily ACL, or someone to finger what in ACL (or ASDF) is causing the problem. Sorry I wasn't more clear.

Here's a reduced test case.

(load (compile-file "al10bug.lisp"))

;;; al10bug.lisp (defparameter *result* nil)

(defun two-optionals (&optional (o1 :default1) (o2 :default2)) (format t "TWO-OPTIONALS ~S ~S~%" o1 o2) (setf *result* (if (equal o1 :parameter) :success :failure)) #'(lambda (x) o2))

(trace two-optionals)

(two-optionals :parameter)

(format t "TEST ~A~%" *result*)

—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org When communists say "to each man according to his needs", remember they established just how low these needs are, by founding concentration camps.

On Sun, Aug 30, 2015 at 6:54 PM, Kevin Layer layer@franz.com wrote:

Great, thanks!

Even slightly more reduced:

(defun two-optionals (&optional (o1 :default1) (o2 :default2)) (format t "TWO-OPTIONALS ~S ~S~%" o1 o2) (lambda (x) o2) (if (equal o1 :parameter) :success :failure))

(trace two-optionals)

(format t "TEST ~A~%" (two-optionals :parameter))

—♯ƒ • François-René ÐVB Rideau •Reflection&Cybernethics• http://fare.tunes.org Does artillery violate the natural rights of the target? I would say: the entire *purpose* of artillery is to violate the natural rights of the target. — Mencius Moldbug