Re: [PATCH 8/8] vm: full exception printing support

On Mon, 2009-07-06 at 20:28 +0200, Tomek Grabiec wrote: > This introduces vm_throwable_print_stack_trace() which generates > a string which is identical to the one that would be printed > if Throwable.printStackTrace() method is executed. > > Signed-off-by: Tomek Grabiec I applied the series. Th

[PATCH 4/8] jit: make signal_new_exception() functions return success/error information

When exception occures during exception allocation then signal_new_exception() functions will return a negative value. If exception occurres after new exception was allocated then new exception is not signalled. Signed-off-by: Tomek Grabiec --- include/jit/exception.h |8 jit/except

[PATCH 3/8] vm: remove trampolines for JNI methods

This introduces a better approach for calling JNI methods. The approach is based on assumption that if native method is not a VM native then it is a JNI method. This lets us to prepare JNI call arguments at compilation time. When method is prepared in vm_method_prepare_jit() it is checked whether

[PATCH 2/8] vm: remove unused ->jit_code field

Signed-off-by: Tomek Grabiec --- include/vm/method.h |2 -- 1 files changed, 0 insertions(+), 2 deletions(-) diff --git a/include/vm/method.h b/include/vm/method.h index ebea698..fd19c16 100644 --- a/include/vm/method.h +++ b/include/vm/method.h @@ -36,8 +36,6 @@ struct vm_method {

Re: [PATCH 1/8] x86: fix emitting branches with target pointing to the containing basic block

2009/7/6 Tomek Grabiec : > When branch target points to the same basic block that contains that > branch then branch insn will be added to the backpatch list because > target_bb->is_emitted field is false yet. The backpatch list for basic block > is browsed before its instructions are emitted. Ther

[PATCH 8/8] vm: full exception printing support

This introduces vm_throwable_print_stack_trace() which generates a string which is identical to the one that would be printed if Throwable.printStackTrace() method is executed. Signed-off-by: Tomek Grabiec --- include/vm/preload.h |9 ++ vm/preload.c | 64 +++- vm/stack

[PATCH 7/8] vm: print method signature on preload failure

Signed-off-by: Tomek Grabiec --- vm/preload.c |3 +++ 1 files changed, 3 insertions(+), 0 deletions(-) diff --git a/vm/preload.c b/vm/preload.c index f930437..9f8de3f 100644 --- a/vm/preload.c +++ b/vm/preload.c @@ -27,6 +27,7 @@ #include +#include "vm/die.h" #include "vm/classloader

[PATCH 1/8] x86: fix emitting branches with target pointing to the containing basic block

When branch target points to the same basic block that contains that branch then branch insn will be added to the backpatch list because target_bb->is_emitted field is false yet. The backpatch list for basic block is browsed before its instructions are emitted. Therefore the branch whose target poi

[PATCH 6/8] vm: implement JNI functions for exception management

Signed-off-by: Tomek Grabiec --- include/vm/jni.h |1 + vm/jni-interface.c | 59 ++- 2 files changed, 54 insertions(+), 6 deletions(-) diff --git a/include/vm/jni.h b/include/vm/jni.h index 8f76ecb..9fbb1a3 100644 --- a/include/vm/jni.h ++

[PATCH 5/8] vm: check in JNI handlers that class objects are instances of java.lang.Class

Signed-off-by: Tomek Grabiec --- vm/jni-interface.c |6 ++ 1 files changed, 6 insertions(+), 0 deletions(-) diff --git a/vm/jni-interface.c b/vm/jni-interface.c index 20a1588..0eff83c 100644 --- a/vm/jni-interface.c +++ b/vm/jni-interface.c @@ -46,6 +46,10 @@ return NULL

Re: [PATCH 1/2] vm: remove unused ->jit_code field

2009/7/6 Tomek Grabiec : > > Signed-off-by: Tomek Grabiec > --- >  include/vm/method.h |    2 -- >  1 files changed, 0 insertions(+), 2 deletions(-) > > diff --git a/include/vm/method.h b/include/vm/method.h > index ebea698..fd19c16 100644 > --- a/include/vm/method.h > +++ b/include/vm/method.h >

[PATCH 2/2] vm: assume that non-VM native methods are JNI native methods

When method is prepared in vm_method_prepare_jit() it is checked whether native method is one of VM native methods and the address of it is stored in the ->vm_native_ptr field. This field is used to perform a quick check for a JNI method in vm_method_is_jni_method() and to omit vm_native_lookup() i

[PATCH 1/2] vm: remove unused ->jit_code field

Signed-off-by: Tomek Grabiec --- include/vm/method.h |2 -- 1 files changed, 0 insertions(+), 2 deletions(-) diff --git a/include/vm/method.h b/include/vm/method.h index ebea698..fd19c16 100644 --- a/include/vm/method.h +++ b/include/vm/method.h @@ -36,8 +36,6 @@ struct vm_method {

Re: [PATCH] x86: fix emitting branches with target pointing to the containing basic block

On Mon, 2009-07-06 at 13:27 +0200, Tomek Grabiec wrote: > When branch target points to the same basic block that contains that > branch then branch insn will be added to the backpatch list because > target_bb->is_emitted field is false yet. The backpatch list for basic block > is browsed before its

[PATCH] x86: fix emitting branches with target pointing to the containing basic block

When branch target points to the same basic block that contains that branch then branch insn will be added to the backpatch list because target_bb->is_emitted field is false yet. The backpatch list for basic block is browsed before its instructions are emitted. Therefore the branch whose target poi

Re: [PATCH 2/2] x86: fix bug in emit_jni_trampoline()

2009/7/6 Pekka Enberg : > On Mon, 2009-07-06 at 01:16 +0200, Tomek Grabiec wrote: >> We cannot overwrite a register which is not saved (ESI) because this >> will lead to a corruption in JIT code. The proper solution is >> to copy the call arguments. >> >> Signed-off-by: Tomek Grabiec > > I'm confu

Re: [PATCH 2/2] x86: fix bug in emit_jni_trampoline()

On Mon, 2009-07-06 at 01:16 +0200, Tomek Grabiec wrote: > We cannot overwrite a register which is not saved (ESI) because this > will lead to a corruption in JIT code. The proper solution is > to copy the call arguments. > > Signed-off-by: Tomek Grabiec I'm confused. What exactly is the problem

Re: [PATCH 1/2] x86: fix bug in invokeinterface and invokevirtual args cleanup

On Mon, 2009-07-06 at 01:16 +0200, Tomek Grabiec wrote: > method->args_count should be used to get the size of arguments, > not nr_args(args_list). This bug led to stack overflow with the > following code: > > public class Test { > public void c(long x) { > } > > public static void ma

Re: [PATCH 2/2] vm: fix native_vmobject_getclass()

On Sun, 2009-07-05 at 22:36 +0200, Tomek Grabiec wrote: > Add missing return after throw_from_native() and put object->class check in > assertion > > Reported-by: Vegard Nossum > Signed-off-by: Tomek Grabiec I applied this but here are some comments: > --- > vm/jato.c |5 - > 1 files

Re: [PATCH 1/2] vm: add license text to jni.c and jni-interface.c

applied -- ___ Jatovm-devel mailing list Jatovm-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/jatovm-devel