Re: [Cp-tools-discuss] gjdoc works again!

[Alex Lancaster]

>>>>> "JS" == Julian Scheid <address@hidden> writes:

JS> ... well at least partially.  I am currently using the Sun VM and
JS> libraries for running the thing, not tested w/Classpath yet.

JS> I get more or less sensible output by using the minimal
JS> Makefile-hack attached below. Please adjust for your system.

JS> Don't expect too much, but I think you can start playing around
JS> with it a little. However it is definitely too early for real
JS> testing or bug reports. There are a good number of known problems
JS> on my list already, I'll enter them into Savannah task manager
JS> tonight.

JS> Note that while our goal should be to get gjdoc+texidoclet
JS> working, it's probably better to test gjdoc using
JS> com.sun.tools.doclets.standard.Standard from Sun's core classes in
JS> the first place, as this is something like a 'reference' doclet in
JS> terms of completeness of API usage.

JS> I plan to finish gnu.classpath.tools.gjdoc.DebugDoclet so that it
JS> retrieves ALL information from the doclet API, sorts it, and
JS> writes it to stdout. (currently, it only retrieves a tiny subset.)
JS> That way, we can simply use diff to verify that gjdoc is
JS> compatible to javadoc and to pinpoint the situations where gjdoc
JS> behaviour deviates.

JS> Until this is implemented however, you should stick to using Sun's
JS> standard doclet in order to evaluate gjdoc's performance and
JS> compatibility. (The Makefile below doesn't reflect this, though.)

Hey Julian!

Great job!  Just checked out the modified sources from CVS.  I
modified your example slightly to work on my system and as you
suggested tested using the "standard" Sun doclet, rather than
texidoclet for the moment. I got the more-or-less correct looking
output for documenting the classpath version of Sun's
"com.sun.javadoc.MethodDoc" interface.  However, if I modified this to
look at the whole package I got a segfault:
 
alex$:~/src/gnu.org/classpath-tools/gjdoc$ make -f Makefile.working run
echo Compiling...
Compiling...
/usr/lib/jdk1.3.1/bin/javac  -d tmp -classpath 
"tmp:../tmp:/usr/lib/jdk1.3.1/lib/tools.jar" gnu/classpath/tools/gjdoc/*.java
/usr/lib/jdk1.3.1/bin/java  -classpath 
"tmp:../tmp:/usr/lib/jdk1.3.1/lib/tools.jar" gnu.classpath.tools.gjdoc.Main 
-doclet com.sun.tools.doclets.standard.Standard com.sun.javadoc -sourcepath 
../../classpath:. -d docs
Loading classes for package com.sun.javadoc ...
make: *** [run] Segmentation fault

Is this a known problem (i.e. something that won't work as yet?)
or should I break out gdb/jdb and fix it?

Here's my modified Makefile:

JAVA_HOME=/usr/lib/jdk1.3.1
RUN_CLASSPATH = tmp:../tmp:$(JAVA_HOME)/lib/tools.jar
#RUN_DOCLET = gnu.classpath.tools.doclets.texinfo.Driver
RUN_DOCLET = com.sun.tools.doclets.standard.Standard
RUN_MAINCLASS = gnu.classpath.tools.gjdoc.Main
#RUN_ARG_CLASSES = com.sun.javadoc.MethodDoc com.sun.javadoc.ClassDoc
RUN_ARG_CLASSES = com.sun.javadoc
RUN_ARG_OPTIONS = -sourcepath ../../classpath:. -d docs

JAVA = $(JAVA_HOME)/bin/java
JAVAC = $(JAVA_HOME)/bin/javac 
#JAVAC = jikes

RELEASE_OPTIONS =
DEBUG_OPTIONS = -Dgnu.classpath.tools.gjdoc.LogLevel=1

create_dirs:
        mkdir tmp
        mkdir docs

all:
        echo Compiling...
        $(JAVAC) -d tmp -classpath "$(RUN_CLASSPATH)" 
gnu/classpath/tools/gjdoc/*.java

run: all
        $(JAVA) $(RELEASE_OPTIONS) -classpath "$(RUN_CLASSPATH)" 
$(RUN_MAINCLASS) -doclet $(RUN_DOCLET) $(RUN_ARG_CLASSES) $(RUN_ARG_OPTIONS)

--