ArmedBear / J

last message in this thread was 2008-12-10 21:05
I've just been trying (guessing how) to run abcl in eclipse, so far
with no success.  Can anyone help?
Or even give me an idea how to run it in jdb?

Don Cohen wrote:
> last message in this thread was 2008-12-10 21:05
> I've just been trying (guessing how) to run abcl in eclipse, so far
> with no success.  Can anyone help?
> Or even give me an idea how to run it in jdb?

[Quick instructions: should be added to the wiki].

The Ant based build contains a target called 'abcl.debug.jpda' which 
starts an ABCL process with a JDWP listener on port 6789 to which you 
may connect 'jdb'.

So, to run ABCL under 'jdb', have two command line shells open.

 From the first run

   cmd1$ ant abcl.debug.jpda

 From the second run

   cmd2$ jdb -sourcepath $ABCL_ROOT/src -classpath 
$ABCL_ROOT/dist/abcl.jar -attach 6789

which will attach the 'jdb' to the running ABCL instance on TCP port 
6789 using the Java Debug Wire Protocol.

As for running under Eclipse/Netbeans, one uses the IDE to create a 
project "on top of" the Ant-based build system.

An example such an integration with a Netbeans ABCL project [can be 
found in the abcl-dynamic-install repository snapshot][1].  To use this 
with the current ABCL tree, copy the 'netbeans-build.xml' file and the 
contents of the 'nbproject' directory into the top-level ABCL directory. 
  Then one may simply open the ABCL project in Netbeans.

[1]: http://abcl-dynamic-install.googlecode.com/svn/trunk/abcl

-- 
"A screaming comes across the sky.  It has happened before, but there is
nothing to compare to it now."

Mark Evenson writes:

 > The Ant based build contains a target called 'abcl.debug.jpda' which 
 > starts an ABCL process with a JDWP listener on port 6789 to which you 
 > may connect 'jdb'.
 > So, to run ABCL under 'jdb', have two command line shells open.
 >  From the first run
 >    cmd1$ ant abcl.debug.jpda
My result:
 $ ant abcl.debug.jpda
 Buildfile: build.xml
 
 abcl.pre-compile:
     [mkdir] Created dir: /tmp/j/build
     [mkdir] Created dir: /tmp/j/build/classes
 
 BUILD FAILED
 /tmp/j/build.xml:195: Class org.apache.tools.ant.taskdefs.condition.Or
 doesn't support the nested "matches" element.
 
 Total time: 1 second

 >  From the second run
I assume that at this point the second won't do me any good.
 >    cmd2$ jdb -sourcepath $ABCL_ROOT/src -classpath 
 > $ABCL_ROOT/dist/abcl.jar -attach 6789
 > which will attach the 'jdb' to the running ABCL instance on TCP port 
 > 6789 using the Java Debug Wire Protocol.
 > 
 > As for running under Eclipse/Netbeans, one uses the IDE to create a 
 > project "on top of" the Ant-based build system.
I tried to create a project on top of the lisp build.
I thought I had done it correctly but I end up with hundreds of errors.
Then I thought I'd try creating one from svn, but I get
 Validate Repository Location operation failed.
I didn't fill in a user/password.

 > An example such an integration with a Netbeans ABCL project [can be 
 > found in the abcl-dynamic-install repository snapshot][1].  To use this 
 > with the current ABCL tree, copy the 'netbeans-build.xml' file and the 
 > contents of the 'nbproject' directory into the top-level ABCL directory. 
 >   Then one may simply open the ABCL project in Netbeans.
 > 
 > [1]: http://abcl-dynamic-install.googlecode.com/svn/trunk/abcl
I'll try to use netbeans while I hope for replies...

Don Cohen writes:
 >  > An example such an integration with a Netbeans ABCL project [can be 
 >  > found in the abcl-dynamic-install repository snapshot][1].  To use this 
 >  > with the current ABCL tree, copy the 'netbeans-build.xml' file and the 
 >  > contents of the 'nbproject' directory into the top-level ABCL directory. 
 >  >   Then one may simply open the ABCL project in Netbeans.
 >  > 
 >  > [1]: http://abcl-dynamic-install.googlecode.com/svn/trunk/abcl
 > I'll try to use netbeans while I hope for replies...
By the top level ABCL directory, I assume you mean j, in my case /tmp/j.
I copied the 5 files into that directory, and now in netbeans 6.5
I do file - open project, find /tmp/j and click on "open project" but
nothing happens.

Next try - back in eclipse, I was able to run (debug) in spite of
errors in the project.  I entered the (load ...) into the console
and it ran to the same conclusion:
 Debugger invoked on condition of type ERROR:
  Out of memory.
I was hoping to see a java stack trace.
I do have "suspend execution on uncaught exceptions" checked.
I guess this is a caught exception?

I see now that there is only one "out of memory" in the source,
so I put a break there.  Here's the stack at the out of memory:

 Thread [interpreter] (Suspended (breakpoint at line 281 in Lisp$1))     
         Lisp$1.execute(LispObject) line: 281    
         Symbol.execute(LispObject) line: 720    
         LispThread.execute(LispObject, LispObject) line: 638    
         top_level_44.execute(LispObject[]) line: 429    
         top_level_44(CompiledFunction).execute(LispObject, LispObject) line: 79 
         Symbol.execute(LispObject, LispObject) line: 733        
         LispThread.execute(LispObject, LispObject, LispObject) line: 659        
         debug_2.execute() line: 68      
         Symbol.execute() line: 708      
         LispThread.execute(LispObject) line: 618        
         debug_3.execute() line: 82      
         Symbol.execute() line: 708      
         LispThread.execute(LispObject) line: 618        
         debug_7.execute(LispObject[]) line: 110 
         CompiledClosure.execute() line: 56      
         LispThread.execute(LispObject) line: 618        
         with_standard_io_syntax_1.execute(LispObject) line: 35  
         Symbol.execute(LispObject) line: 720    
         LispThread.execute(LispObject, LispObject) line: 638    
         debug_6.execute(LispObject) line: 110   
         Symbol.execute(LispObject) line: 720    
         LispThread.execute(LispObject, LispObject) line: 638    
         signal_2._execute(LispObject, LispObject) line: 65      
         signal_2(Primitive1R).execute(LispObject) line: 87      
         Symbol.execute(LispObject) line: 720    
         Lisp.error(LispObject) line: 338        
         Lisp$1.execute(LispObject) line: 281    
         Symbol.execute(LispObject) line: 720    
         LispThread.execute(LispObject, LispObject) line: 638    
         top_level_44.execute(LispObject[]) line: 429    
         top_level_44(CompiledFunction).execute() line: 63       
         Symbol.execute() line: 708      
         LispThread.execute(LispObject) line: 618        
         top_level_45.execute() line: 439        
         LispThread.execute(LispObject) line: 618        
         Interpreter.run() line: 373     
         Main$1.run() line: 50   
         Thread.run() line: 619  

How can I figure out what's going on here?
Is there some way to call lisp PRINT on the arguments?
When I put the cursor over some things like arg in
  return function.execute(arg);  (line 720 - several occurrences above)
I see things like 
 Cons 
 CompiledClosure 
 LispError 
 LispError 
 Cons

BTW, the error is coming near the end of a call to compile-file.

 Thread [interpreter] (Suspended (breakpoint at line 281 in Lisp$1))     
         Lisp$1.execute(LispObject) line: 281    
         Symbol.execute(LispObject) line: 720    
         LispThread.execute(LispObject, LispObject) line: 638    
         top_level_44.execute(LispObject[]) line: 429    
         top_level_44(CompiledFunction).execute(LispObject, LispObject) line: 79 
         Symbol.execute(LispObject, LispObject) line: 733        
The two lisp objects here both seem to be streams.
Can that be right?

         LispThread.execute(LispObject, LispObject, LispObject) line: 659        
         debug_2.execute() line: 68      
I just noticed that this seems to be a lisp file, 
(defun internal-debug ()
  (if (fboundp 'tpl::repl)  ;; <= line 68

Could all above be trying to determine whether tpl::repl is fboundp ?

         Symbol.execute() line: 708      
         LispThread.execute(LispObject) line: 618        
         debug_3.execute() line: 82      
(defun debug-loop ()
  (let ((*debug-level* (1+ *debug-level*)))
    (show-restarts (compute-restarts) *debug-io*)
    (internal-debug))) ; <= line 82

         Symbol.execute() line: 708      
         LispThread.execute(LispObject) line: 618        
         debug_7.execute(LispObject[]) line: 110 

    (unless (fboundp 'tpl::repl)
      (quit)) ; <= line 110
so we're trying to quit?

         CompiledClosure.execute() line: 56      
         LispThread.execute(LispObject) line: 618        
         with_standard_io_syntax_1.execute(LispObject) line: 35  
         Symbol.execute(LispObject) line: 720    
         LispThread.execute(LispObject, LispObject) line: 638    
         debug_6.execute(LispObject) line: 110   
looks just like debug_7 above

         Symbol.execute(LispObject) line: 720    
         LispThread.execute(LispObject, LispObject) line: 638    
         signal_2._execute(LispObject, LispObject) line: 65      
(defun error (datum &rest arguments)
;; line 65 below
  (let ((condition (coerce-to-condition datum arguments 'simple-error 'error)))

         signal_2(Primitive1R).execute(LispObject) line: 87      
        return _execute(arg, NIL);
arg above displays as UndefinedFunction

I also see line numbers past the end of the file top-level.lisp.

At the break I tried (step (compile-file* "treegen")) in hope of
seeing some intermediate states, but it got back to the error
immediately. 

I hope this does you more good than it does me.

Don Cohen wrote:
[…]
>  > I'll try to use netbeans while I hope for replies...
> By the top level ABCL directory, I assume you mean j, in my case /tmp/j.
> I copied the 5 files into that directory, and now in netbeans 6.5
> I do file - open project, find /tmp/j and click on "open project" but
> nothing happens.

Roughly that order of instructions to run ABCL under Netbeans worked 
last time I tried that for me, but I will check again.  There is a 
'abcl/nbproject/private' that is not propagated the public tree, so it 
is highly likely that I have unresolved runnability issues.

I will try to reproduce a virgin build from my end.

Mark <evenson@...>

-- 
"A screaming comes across the sky.  It has happened before, but there is
nothing to compare to it now."

Don Cohen wrote:
> Mark Evenson writes:
> 
>  > The Ant based build contains a target called 'abcl.debug.jpda' which 
>  > starts an ABCL process with a JDWP listener on port 6789 to which you 
>  > may connect 'jdb'.
>  > So, to run ABCL under 'jdb', have two command line shells open.
>  >  From the first run
>  >    cmd1$ ant abcl.debug.jpda
> My result:
>  $ ant abcl.debug.jpda
>  Buildfile: build.xml
>  
>  abcl.pre-compile:
>      [mkdir] Created dir: /tmp/j/build
>      [mkdir] Created dir: /tmp/j/build/classes
>  
>  BUILD FAILED
>  /tmp/j/build.xml:195: Class org.apache.tools.ant.taskdefs.condition.Or
>  doesn't support the nested "matches" element.

It might be that your version of Ant is too old:  we kind of require Ant 
1.7 or later at this point.  And unfortunately, we do not refuse to try 
to execute if the build.xml file is invoked from an earlier version of Ant.

You could check which Ant version you are running via:

   cmd$ ant -version

-- 
"A screaming comes across the sky.  It has happened before, but there is
nothing to compare to it now."

Don Cohen wrote:
> Don Cohen writes:
>  >  > An example such an integration with a Netbeans ABCL project [can be 
>  >  > found in the abcl-dynamic-install repository snapshot][1].  To use this 
>  >  > with the current ABCL tree, copy the 'netbeans-build.xml' file and the 
>  >  > contents of the 'nbproject' directory into the top-level ABCL directory. 
>  >  >   Then one may simply open the ABCL project in Netbeans.
>  >  > 
>  >  > [1]: http://abcl-dynamic-install.googlecode.com/svn/trunk/abcl
>  > I'll try to use netbeans while I hope for replies...
> By the top level ABCL directory, I assume you mean j, in my case /tmp/j.
> I copied the 5 files into that directory, and now in netbeans 6.5
> I do file - open project, find /tmp/j and click on "open project" but
> nothing happens.

By "top level directory", I mean a Ant buildable ABCL directory 
containing the 'build.xml' and the 'src' subdirectory with ABCL sources. 
  If you check ABCL out from the [common-lisp.net svn repo][3], the 
top-level directory may very well be named 'j' (if you extract it from 
an ABCL source distribution the "top level directory" will be named 
something like 'abcl-src-0.12.0').

Into this directory graft

   netbeans-build.xml
   nbproject/build-impl.xml
   nbproject/genfiles.properties
   nbproject/project.properties
   nbproject/project.xml

from [the ABCL dynamic install repository][1].

I [just updated these files][2] to fix problems with actually building 
ABCL in Netbeans, but you should have been at least able to open the 'j' 
directory in your case.

[1]: http://abcl-dynamic-install.googlecode.com/svn/trunk/abcl

[2]: http://code.google.com/p/abcl-dynamic-install/source/detail?r=211

[3]: svn://common-lisp.net/project/armedbear/svn/trunk/j

-- 
"A screaming comes across the sky.  It has happened before, but there is
nothing to compare to it now."

I've been hoping that you'd read ahead and see that I succeeded in
running under eclipse.  So all of this is no longer on the critical
path toward compiling ap5.

But if you're still interested in finding ways to run abcl under
various debuggers then I'm willing to try various things.  But I 
think we have to make sure that I'm interpreting your words as
you intend.  My use of 
 [1]: http://abcl-dynamic-install.googlecode.com/svn/trunk/abcl
was as follows:
I went to the web page, followed links to 5 pages and did save-as
into /tmp/j, so I now have the following in my emacs dired sorted by
time: 

 /tmp/j:
  total 6428
  drwxr-xr-x  11 root root    4096 Dec 18 08:52 .
  -rw-r--r--   1 root root     576 Dec 18 08:52 project.xml
  -rw-r--r--   1 root root    2062 Dec 18 08:51 project.properties
  -rw-r--r--   1 root root     617 Dec 18 08:51 genfiles.properties
  -rw-r--r--   1 root root   32750 Dec 18 08:51 build-impl.xml
  -rw-r--r--   1 root root     752 Dec 18 08:50 netbeans-build.xml
  drwxrwxrwt 106 root root   69632 Dec 18 08:46 ..
  drwxr-xr-x   3 root root    4096 Dec 18 07:40 build
  drwxr-xr-x   7 root root    4096 Dec 17 15:58 src
  drwxr-xr-x   2 root root    4096 Dec 17 15:58 .settings
 ...
I then run netbeans:
 [2008-12-18 09:06:22 root@... /usr/local/netbeans-6.5/bin]
 $ ./netbeans
click on File - open project
This brings up a new window with a file chooser.
If I choose
 /tmp/j
then the "open project" button does nothing.
It looks like it's looking for some files that it's not finding.
The right side looks like it's disabled.  It contains
 project name _________
 [] open as main project
 [] open required projects
In fact, I try selecting other locations under /tmp/j and none of 
them cause this to change appearance, or cause open project to
do anything.
So perhaps those extra files are supposed to be somewhere else, like
in src ?  Copying them to src doesn't seem to help.

So I think that I'm just not understanding your directions here.

The use of svn was another unrelated attempt.
That was in eclipse.  If you want to try to figure out how to make
that work for me, I'm willing to try more detailed directions for
that too.

 > I just did a fresh checkout from SVN that verifies that [the 
 > abcl-dynamic-install repository][1] works with Netbeans 6.5 on OSX 
 > 10.5.6 by simply opening the 'abcl' directory as a project.
This sounds different from what I did.
If you want me to try it, send instructions.
It sounds like there's one svn command line followed by another to start
netbeans and then some instructions (similar to above) in netbeans.

> I went to the web page, followed links to 5 pages and did save-as
> into /tmp/j, so I now have the following in my emacs dired sorted by
> time:
>
>  /tmp/j:
>  total 6428
>  drwxr-xr-x  11 root root    4096 Dec 18 08:52 .
>  -rw-r--r--   1 root root     576 Dec 18 08:52 project.xml
>  -rw-r--r--   1 root root    2062 Dec 18 08:51 project.properties
>  -rw-r--r--   1 root root     617 Dec 18 08:51 genfiles.properties
>  -rw-r--r--   1 root root   32750 Dec 18 08:51 build-impl.xml
>  -rw-r--r--   1 root root     752 Dec 18 08:50 netbeans-build.xml

For what I know of netbeans, netbeans-build.xml should be named
build.xml and the other 4 files should be in a directory named
"nbproject" (at the same level as build.xml). The open dialog of
netbeans recognizes when a directory contains a project and shows it
appropriately (e.g. it changes the icon and enables the buttons).

hth,
Alessio

>    netbeans-build.xml
 >    nbproject/build-impl.xml
 >    nbproject/genfiles.properties
 >    nbproject/project.properties
 >    nbproject/project.xml
Ok, this was the problem.
I interpreted your original directions as copying nbproject/x into 
/tmp/j/x.  Now that I put it in /tmp/j/nbproject/x the open project
works.
I don't know how to use netbeans, but it looks like there are a bunch
of errors.  However, this is no surprise since I saw the same thing in
eclipse, and there I was still able to run it.

Anyhow, I think this particular problem is now solved.
Since I have a better idea of what I'm doing in eclipse, I'll stick
with that for now.

On Fri, Dec 19, 2008 at 4:58 PM, Don Cohen
<don-sourceforge-xxz@...> wrote:
>
>  >    netbeans-build.xml
>  >    nbproject/build-impl.xml
>  >    nbproject/genfiles.properties
>  >    nbproject/project.properties
>  >    nbproject/project.xml
> Ok, this was the problem.
> I interpreted your original directions as copying nbproject/x into
> /tmp/j/x.  Now that I put it in /tmp/j/nbproject/x the open project
> works.
> I don't know how to use netbeans, but it looks like there are a bunch
> of errors.  However, this is no surprise since I saw the same thing in
> eclipse, and there I was still able to run it.
>
> Anyhow, I think this particular problem is now solved.
> Since I have a better idea of what I'm doing in eclipse, I'll stick
> with that for now.

I do prefer Eclipse myself, in fact I'm using it to develop on ABCL.
NetBeans is just plain crap. Now let's start flame wars! :D

...I was joking, if that wasn't clear (not about NB being crap...
about the flame wars :D)

A.

Alessio Stalla wrote:

[More explicit instructions for using ABCL with Netbeans]

> For what I know of netbeans, netbeans-build.xml should be named
> build.xml […]

Actually, the current 'build.xml' attempts to optionally include (see 
line 540) 'netbeans-build.xml' if it is present, so just copying that 
file over is what is needed (no need to clobber the existing 
'build.xml').  Such coexistence is part of the reason that all the ABCL 
Ant build targets and variables are prefixed with 'abcl.' or 'j.' with 
the other part of the reason is so that the old Ant buildfile for J can 
also be used in such a manner.


-- 
"A screaming comes across the sky.  It has happened before, but there is
nothing to compare to it now."

I was hoping that we could get past the part about how to run abcl
from a java debugger and into the question of what's causing the
out-of-memory at the end of compile-file.  I hope I've provided enough
information for you to reproduce the problem.  I was also hoping that
just seeing the backtrace might give you an idea of what was going on.
It does seem to be internal to compile-file, at the end, after output
of 
; Compilation unit finished
;   The following functions were used but not defined:
 ...
so in order to get much further I'd have to find and understand the
code for compile-file.  I'm guessing (hoping) that you already know
something about that.
Let me know what you think is the best way to proceed and what you
want me to do.

On Sat, Dec 20, 2008 at 6:11 PM, Don Cohen
<don-sourceforge-xxz@...> wrote:
>
> I was hoping that we could get past the part about how to run abcl
> from a java debugger and into the question of what's causing the
> out-of-memory at the end of compile-file.  I hope I've provided enough
> information for you to reproduce the problem.

Hi Don,

I'm sorry to see this isn't going the way you hoped. Probably you did
provide enough information. However, I'm working at some other parts
of the system, creating new infrastructure to (hopefully) be able to
fix some of the ansi compliance tests. This is tedious work and
distraction to your problem would probably leave me in a very sloppy
state. You'll understand that that's definitely undesirable.

Also, Ville has gone on holidays, meaning that we're halved in man power.

>  I was also hoping that
> just seeing the backtrace might give you an idea of what was going on.

Unfortunately, that didn't happen. It's hard to recognise anything
from Java stack traces in ABCL: the code it generates is all called
"execute", meaning that there's very little we can actually see from
the trace, normally. The traces do help, but only after we gain more
insight into the problem.

> It does seem to be internal to compile-file, at the end, after output
> of
> ; Compilation unit finished
> ;   The following functions were used but not defined:
>  ...

Can you put on a download site somewhere the full directory you're
building from? I can have a go. It's possible that compile-file uses
undefined functions, but we'd expect those functions to be defined
elsewhere in the system. So, if they're not, that's alarming.

If I can download your build setup, I'll be able to diagnose your
problem the fastest (hopefully).

> so in order to get much further I'd have to find and understand the
> code for compile-file.  I'm guessing (hoping) that you already know
> something about that.

:-) Yes, some: I've become acquainted with the code structure for the
last 6 months quite a bit. However, the code in compile-file hasn't
been touched/studied much by me. That's no problem though: we can work
at it together.

> Let me know what you think is the best way to proceed and what you
> want me to do.

Would it be possible to contact me through IRC somewhere during the
european evening? (9 or 10 pm CET?)

I'm usually hanging around on IRC a lot of time (but not all the time)
during the evening at #abcl on FreeNode
(irc://irc.freenode.net/#abcl).


Bye,

Erik.

Erik Huelsmann writes:

 > I'm sorry to see this isn't going the way you hoped.
I wasn't trying to rush you, just direct attention away from the
problem that was no longer on the critical path (how to run in some
debugger) and toward the one that is. 

 > Can you put on a download site somewhere the full directory you're
 > building from? I can have a go. It's possible that compile-file uses
 > undefined functions, but we'd expect those functions to be defined
 > elsewhere in the system. So, if they're not, that's alarming.
I guess I could do that, but I think it's easier to start from
 http://ap5.com/ap5-2006.zip 
and apply the few small diffs I sent, along with the small build
script. 

 > Would it be possible to contact me through IRC somewhere during the
 > european evening? (9 or 10 pm CET?)
I'm trying ...
 > I'm usually hanging around on IRC a lot of time (but not all the time)
 > during the evening at #abcl on FreeNode
 > (irc://irc.freenode.net/#abcl).
I may not be doing it right, having no experience with this irc
client.  I seem to now be don@..., if that helps.
I guess you're not logged in.
14:00PST +8 = 2200 UT, I think.

On Tue, Dec 23, 2008 at 11:07 PM, Don Cohen
<don-sourceforge-xxz@...> wrote:
> Erik Huelsmann writes:
>
>  > I'm sorry to see this isn't going the way you hoped.
> I wasn't trying to rush you, just direct attention away from the
> problem that was no longer on the critical path (how to run in some
> debugger) and toward the one that is.
>
>  > Can you put on a download site somewhere the full directory you're
>  > building from? I can have a go. It's possible that compile-file uses
>  > undefined functions, but we'd expect those functions to be defined
>  > elsewhere in the system. So, if they're not, that's alarming.
> I guess I could do that, but I think it's easier to start from
>  http://ap5.com/ap5-2006.zip
> and apply the few small diffs I sent, along with the small build
> script.

Ok, I'll do that then.

>  > Would it be possible to contact me through IRC somewhere during the
>  > european evening? (9 or 10 pm CET?)
> I'm trying ...

>  > I'm usually hanging around on IRC a lot of time (but not all the time)
>  > during the evening at #abcl on FreeNode
>  > (irc://irc.freenode.net/#abcl).

> I may not be doing it right, having no experience with this irc
> client.  I seem to now be don@..., if that helps.
> I guess you're not logged in.
> 14:00PST +8 = 2200 UT, I think.

That's right. Meaning that it was 23:00 CET (CET being UTC + 1). I was
there, but I guess we missed each other. My nick is ehu or ehu`, but
if you have access to GMail chat, then it might be easier to get in
contact. Or, if you have Firefox, when you install Chatzilla, clicking
on the link I provided should get you directly into the channel.
Either way, I'll probably be online tonight between 13:00 and 15:00PST
(being 20:00 and 22:00 CET).

Hope that helps. I'll be installing the packages mentioned first. The
problem Brian Jones is reporting is closely related (I expect) to the
issue I'm trying to resolve already, so I'm working on that first, but
with the holidays coming, I might have time to look into both issues.


Bye,

Erik.