CEDET

>>> ludo@... (Ludovic =?iso-8859-1?Q?Court=E8s?=) seems to think that:
>Hi,
>
>ludo@... (Ludovic Courtès) writes:
>
>> Hi,
>>
>> EIEIO's test suite from CEDET 1.0pre4 fails:
>>
>>   for loadpath in . ../common/ ../speedbar/ ; do \
>>              echo "(add-to-list 'load-path \"$loadpath\")" >> test-compile-script; \
>>           done;
>>   "emacs" -batch --no-site-file -l test-compile-script -f batch-byte-compile eieio-tests.el eieio-test-methodinvoke.el
>>   !! File eieio uses old-style backquotes !!
>>   Class allocated slots do not need :initarg
>>
>>   In toplevel form:
>>   eieio-tests.el:474:1:Error: Symbol's value as variable is void: class-typep-var
>>
>> This is with GNU Emacs 22.3.
>
>FWIW the issue still shows up with CEDET 1.0pre6, followed by another
>one:
>
>  In toplevel form:
>  eieio-tests.el:521:1:Error: Symbol's value as variable is void: class-typep-var
>  Wrote /tmp/nix-build-hajhxyz4l5pd8lsl89df3ina4fyd9vfi-cedet-1.0pre6.drv-0/cedet-1.0pre6/eieio/eieio-test-methodinvoke.elc
>
>  In end of data:
>  eieio-perftest.el:149:1:Warning: the function `semantic-elapsed-time' is not
>      known to be defined.
  [ ... ]

Hi,

  In my previous reply, there is one thing I hadn't noticed before,
which is that you must have typed "make test" in the eieio directory
to do this.

  The command "make test" in the EIEIO directory is not actually
supposed to be run.  I hadn't realized that is how you got into this
state.  I thought you were running "make utest" from the cedet
directory, not "make test" from the eieio directory.  The "make test"
command attempts to compile the tests file, which cannot be compiled.

  Was there something that made you think this should be run to test
something?

  The fact that it exists is a side effect of the makefile generator
being used, but it should not be run during normal compilation.

Eric

-- 
          Eric Ludlam:                       eric@...
   Siege: http://www.siege-engine.com          Emacs: http://cedet.sourceforge.net

>>> ludo@... (Ludovic =?iso-8859-1?Q?Court=E8s?=) seems to think that:
>Hi,
>
>ludo@... (Ludovic Courtès) writes:
>
>> Hi,
>>
>> EIEIO's test suite from CEDET 1.0pre4 fails:
>>
>>   for loadpath in . ../common/ ../speedbar/ ; do \
>>              echo "(add-to-list 'load-path \"$loadpath\")" >> test-compile-script; \
>>           done;
>>   "emacs" -batch --no-site-file -l test-compile-script -f batch-byte-compile eieio-tests.el eieio-test-methodinvoke.el
>>   !! File eieio uses old-style backquotes !!
>>   Class allocated slots do not need :initarg
>>
>>   In toplevel form:
>>   eieio-tests.el:474:1:Error: Symbol's value as variable is void: class-typep-var
>>
>> This is with GNU Emacs 22.3.
>
>FWIW the issue still shows up with CEDET 1.0pre6, followed by another
>one:
>
>  In toplevel form:
>  eieio-tests.el:521:1:Error: Symbol's value as variable is void: class-typep-var
>  Wrote /tmp/nix-build-hajhxyz4l5pd8lsl89df3ina4fyd9vfi-cedet-1.0pre6.drv-0/cedet-1.0pre6/eieio/eieio-test-methodinvoke.elc
>
>  In end of data:
>  eieio-perftest.el:149:1:Warning: the function `semantic-elapsed-time' is not
>      known to be defined.
>
  [ ... ]

Hi,

  I was not able to reproduce this issue.  I've run these tests on
several Emacsen, though not 22.3.  I did use 22.1 though.  It might be
as simple as needing a (require 'cl-macs) somewhere.

Eric

-- 
          Eric Ludlam:                       eric@...
   Siege: http://www.siege-engine.com          Emacs: http://cedet.sourceforge.net

Hi,

"Eric M. Ludlam" <eric@...> writes:

>   In my previous reply, there is one thing I hadn't noticed before,
> which is that you must have typed "make test" in the eieio directory
> to do this.
>
>   The command "make test" in the EIEIO directory is not actually
> supposed to be run.  I hadn't realized that is how you got into this
> state.  I thought you were running "make utest" from the cedet
> directory, not "make test" from the eieio directory.  The "make test"
> command attempts to compile the tests file, which cannot be compiled.
>
>   Was there something that made you think this should be run to test
> something?

I think I initially looked for a `check' target (as with GNU
standard compliant makefiles), then tried `test', and finally found out
there was such a rule in sub-directory makefiles.

Anyway, "make utest" in the top-level directory works like a charm.
(I'm actually working on a package for NixOS.)

Thanks!

Ludo'.