Re: [j-devel] compiling files with dots in their names & building abcl with itself

[Peter G. <pe...@ar...>]

On Wed, 31 Aug 2005 at 14:11:09 +0200, Andras Simon wrote:
>
> When compiling a file with a dot in its name, I get
>
> ; Compiling /home/simon/java/jfiles/MidiClasses/bug.bug.cl ...
> ; (IN-PACKAGE :CL-USER)
> ; (DEFUN FOO ...)
> java.lang.ClassFormatError: Illegal class name
> "org/armedbear/lisp/bug.bug_1" in
>   class file <Unknown>
>          at java.lang.ClassLoader.defineClass1(Native Method)
>          at java.lang.ClassLoader.defineClass(ClassLoader.java:620)
>          at java.lang.ClassLoader.defineClass(ClassLoader.java:465)
>          at org.armedbear.lisp.JavaClassLoader.loadClassFromByteArray(JavaClassLoader.java:96)
>          at org.armedbear.lisp.Lisp.loadCompiledFunction(Lisp.java:1021)
>          at org.armedbear.lisp.Lisp.loadCompiledFunction(Lisp.java:959)
> ...
>
> The following patch seems to work (though I'm not sure if this is the
> right thing to do):
>
> Index: src/org/armedbear/lisp/compile-file.lisp
> ===================================================================
> RCS file: /cvsroot/armedbear-j/j/j/src/org/armedbear/lisp/compile-file.lisp,v
> retrieving revision 1.116
> diff -u -r1.116 compile-file.lisp
> --- src/org/armedbear/lisp/compile-file.lisp	27 Aug 2005 13:36:41
> -0000	1.116 +++ src/org/armedbear/lisp/compile-file.lisp	31
> Aug 2005 11:05:11 -0000 @@ -30,7 +30,7 @@
>   (declaim (ftype (function () t) next-classfile-name))
>   (defun next-classfile-name ()
>     (let ((name (%format nil "~A-~D"
> -                       (pathname-name *output-file-pathname*)
> +                       (substitute #\_ #\. (pathname-name
> *output-file-pathname*))
>                          (incf *class-number*))))
>       (namestring (merge-pathnames (make-pathname :name name :type "cls")
>                                    *output-file-pathname*))))
> @@ -439,7 +439,7 @@
>                   (pathnames ()))
>               (dotimes (i *class-number*)
>                 (let* ((file-namestring (%format nil "~A-~D.cls"
> -                                               (pathname-name output-file)
> +                                               (substitute #\_ #\.
> (pathname-name output-file))
>                                                  (1+ i)))
>                        (pathname (merge-pathnames file-namestring output-file)))
>                   (when (probe-file pathname)

I've committed this. It's certainly the right thing to do in the spirit
of the current code. (The current code in that area could stand a bit
of reorganization, but that's a separate issue.)

Thanks!

> It seems unrelated (ie., same problem with unpatched compile-file),
> but I do get the following when building abcl with itself, at jar
> time:
>
> 529.233 seconds real time
> 79599260 cons cells
>
> #
> # An unexpected error has been detected by HotSpot Virtual Machine:
> #
> #  SIGBUS (0x7) at pc=0xb7844a62, pid=26061, tid=3086935744
> #
> # Java VM: Java HotSpot(TM) Client VM (1.5.0-b64 mixed mode)
> # Problematic frame:
> # C  [libzip.so+0xfa62]
> #
> # An error report file with more information is saved as hs_err_pid26061.log
> #
> # If you would like to submit a bug report, please visit:
> #   http://java.sun.com/webapps/bugreport/crash.jsp
> #

I've seen similar things once in a while since abcl started using the
Java zip API. I think it's a Java bug; application code that passes the
verifier shouldn't be able to crash the JVM.

-Peter