From: <as...@ma...> - 2003-12-13 01:29:25
|
All errors result in ABL dying with java.lang.NoSuchFieldError. I don't know how recent this problem is (it disappears by reverting to the 2003-12-11 version), because I only realized now that that may be the reason why ilisp hangs (and the Java process disappears) on startup. (Ilisp always starts with a Lisp error.) For a day or so I thought it was because of the new ARGLIST (which finds the arglist of special operators by retrieving their macro version), and that's why I didn't commit it, even though it worked outside of ilisp. Anyway, this is a bug, never mind ARGLIST or ilisp. Andras |
From: Peter G. <pe...@ar...> - 2003-12-13 01:47:42
|
On Sat, 13 Dec 2003 at 03:28:06 +0100, Andr=E1s_Simon wrote: > All errors result in ABL dying with java.lang.NoSuchFieldError. = I'm not seeing this problem. NoSuchFieldError often means that the source is out of sync somehow (a field has been deleted from some .java file, but the .class files that reference it haven't all been rebuilt). Does ABL still die if you resync with SourceForge CVS and do "make clean classes" (or "ant clean compile")? I've touched almost all the .java files today, reworking the error system a bit, so I'm not totally surprised that something might be out of sync. One noticeable change is that I've replaced all instances (actually all but three, I think, which need to stay) of: throw new ConditionThrowable(new <your condition here>); with: signal(new <your condition here>); in preparation for doing HANDER-BIND, HANDLER-CASE, and SIGNAL the right way. Right now error messages aren't being printed out in a useful format: CL-USER(5): zork Error: #<UNBOUND-VARIABLE @ #x10e3293>. BREAK called. "#<UNBOUND-VARIABLE @ #x10e3293>" being not very informative or friendly. But the current source in my tree (which I synced with CVS within the last half hour or so) has no new failures on the ANSI test suite, so I think it's basically OK. -Peter |
From: Peter G. <pe...@ar...> - 2003-12-13 02:04:35
|
On Fri, 12 Dec 2003 at 17:48:43 -0800, Peter Graves wrote: > Right now error messages aren't being printed out in a useful format: > > CL-USER(5): zork > Error: #<UNBOUND-VARIABLE @ #x10e3293>. > BREAK called. > > "#<UNBOUND-VARIABLE @ #x10e3293>" being not very informative or > friendly. I just checked in a new version of top-level.lisp which fixes this. Now we get: CL-USER(5): zork Error: the variable ZORK has no value. BREAK called. Much better... ;) I also just re-ran my normal working subset of the ANSI test suite, with no unexpected failures. -Peter |
From: <as...@ma...> - 2003-12-13 02:24:52
|
On Fri, 12 Dec 2003, Peter Graves wrote: > On Sat, 13 Dec 2003 at 03:28:06 +0100, Andr=E1s_Simon wrote: > > All errors result in ABL dying with java.lang.NoSuchFieldError. = > > > I'm not seeing this problem. You're right, I was out of sync. Thanks! Andras |