jamvm-general — General discussion of the Jam Virtual Machine.

[Jamvm-general] IcedTea 2.5.3: Avian, JamVM, Cacao: Implement JVM_FindClassFromCaller OpenJDK 8015256: Better class accessibility

[Xerxes R. <xe...@za...>]

Den 2014-10-20 22:30, Matthias Klose skrev:
>
>   - Did you try to build/run these with
>     the IcedTea 2.5.3 update? At least Cacao and JamVM fail.
>
>

I received the IcedTea 2.5.3 security update during the week and indeed 
all alternative JVM broke.
http://blog.fuseyism.com/index.php/2014/10/15/security-icedtea-2-5-3-for-openjdk-7-released/

The OpenJDK source tree revealed the following information:
8015256: Better class accessibility
Summary: Improve protection domain check in forName()
http://hg.openjdk.java.net/jdk7u/jdk7u/hotspot/rev/61d1e75e0a58
http://hg.openjdk.java.net/jdk7u/jdk7u/jdk/rev/16cd2826a58f

JVM_FindClassFromCaller appears to work quite similar to 
JVM_FindClassFromClassLoader with the twist that the protection domain 
that belongs to the caller class argument shall be used during the 
lookup of the class. But there is no specification or unit-test in 
OpenJDK documenting the desired effect, if someone have a specification 
or test at hand for how JVM_FindClassFromCaller "security" shall behave 
then i would like to see it.



I spent some time yesterday looking into it and have filed patches to 
Avian, JamVM and Cacao upstream to make the JVM's compatible with the 
"security" update.
Pull request links and commits below:

JamVM: http://sourceforge.net/p/jamvm/mailman/message/32972760/
https://github.com/xranby/jamvm/commit/81f280b4fad847bc393ee4732c23aae9adccb327

CACAO JVM: 
https://bitbucket.org/cacaovm/cacao-staging/pull-request/147/implement-jvm_findclassfromcaller-openjdk/
https://bitbucket.org/xranby/cacao-staging/commits/ec6bd33b3e927738d1353e6e639e76f74d55635f

Avian: https://github.com/ReadyTalk/avian/pull/360
https://github.com/xranby/avian/commit/2e5990a6b0c35934b99a0a776762fab8f643599b

Cheers
Xerxes

Re: [Jamvm-general] Error building JAMVM on Mac

[Robert L. <rob...@gm...>]

Great you've got it to build!  Obviously, as a 32-bit executable you
can't use a heap of more than a couple of GB.  BTW, the option is
--enable-ffi, not --enable-libffi.

Rob.

On 27 October 2014 00:22, Anitha B Gollamudi <ani...@gm...> wrote:
> On 26 October 2014 11:48, Robert Lougher <rob...@gm...> wrote:
>> Yes it is active.  Unfortunately there's not a lot of help that I can
>> give.  Macs used to be my main computers so they were well supported
>> (under Linux and Mac OS X).  However, I no longer use Macs (my last
>> Mac is from 2006 running OS X 10.5).  Luckily it appears that JamVM
>> continued to be buildable on OS X at least until a year or two ago
>> (judging by feedback) but obviously this is no longer the case.
>> Without a current Mac there's not a lot I can do, so I'll probably
>> remove support entirely from the next version.
>>
>> In case this is of help, I can see the following from your original post.
>>
>> 1) You no longer have gcc on Mac OS X.  The C compiler is actually
>> Clang/LLVM, which is masquerading as gcc.  I have never built JamVM
>> using LLVM as it uses several gcc features.  I no idea if JamVM
>> compiles with LLVM.  Even if it does compile, it may not work.
>>
>> 2) By default, it is building 64-bit executables.  In the past, even
>> on 64-bit machines, gcc on Mac OS X defaulted to 32-bit.  The error
>> messages you are seeing is because JamVM by default (on Mac OS X)
>> assumes it is being built for 32-bit.
>>
>> You could try configuring JamVM to use libffi, e.g. ./configure
>> --enable-ffi.  As far as I'm aware, Mac OS X should still support
>> libffi.  If this works you will have a 64-bit executable.
>>
>> Alternatively you could see if there is a flag to LLVM/Clang to force
>> it to build in 32-bit mode.
>
> Rob,
>
> Finally, I am able to build 1.5.4 version forcing clang to build
>
> [1] Using -m32.
> [2] Removing "interp_cflags=-fno-reorder-blocks" (Clang does not like
> this option)
>
> Bare invocation worked so presumably everything should work though I
> haven't run tests. As a side note --enable-libffi did not work either.
> Configure (clang) complained that it does not know what the option
> meant.
>
> Thanks for your prompt help
>
> -Anitha

[Jamvm-general] [PATCH] Array object modifications

[<chr...@gm...>]

From: Christopher Friedt <chr...@gm...>

This modification to array objects allows for:
1) the classic case where array data is allocated with the array object,
contiguously
2) a new application that allows setting the array data to an arbitrary
pointer (e.g. mmapped file)

---
 src/alloc.c                | 41 +++++++++++++++++++++++++++++++++++------
 src/interp/engine/interp.c | 17 +++++++++--------
 src/jam.h                  | 12 +++++++++++-
 3 files changed, 55 insertions(+), 15 deletions(-)

diff --git a/src/alloc.c b/src/alloc.c
index d12abb6..8b31a69 100644
--- a/src/alloc.c
+++ b/src/alloc.c
@@ -2114,7 +2114,7 @@ Object *allocObject(Class *class) {
 }
     
 Object *allocArray(Class *class, int size, int el_size) {
-    Object *ob;
+    ArrayObject *ob;
 
     /* Special check to protect against integer overflow */
     if(size > (INT_MAX - sizeof(uintptr_t) - sizeof(Object)) / el_size) {
@@ -2122,16 +2122,47 @@ Object *allocArray(Class *class, int size, int el_size) {
         return NULL;
     }
 
-    ob = gcMalloc(size * el_size + sizeof(uintptr_t) + sizeof(Object));
+    ob = gcMalloc(size * el_size + sizeof(ArrayObject));
 
     if(ob != NULL) {
         ob->class = class;
-        ARRAY_LEN(ob) = size;
+        ob->size  = size;
+        if ( size > 0 ) {
+            ob->data = ob->contig_data;
+        }
         TRACE_ALLOC("<ALLOC: allocated %s array object @%p>\n",
                     CLASS_CB(class)->name, ob);
     }
 
-    return ob;
+    return (Object *)ob;
+}
+
+static char *array_names[] = { "[Z", "[C", "[F", "[D",
+	"[B", "[S", "[I", "[J", };
+
+Object *allocTypeArrayFromClassName(const char *className, int size) {
+	Object *r = NULL;
+	int type;
+	char t;
+
+	if ( NULL == className || strlen(className) != 2 || *className != '[' )
+		goto out;
+	t = *(className+1);
+
+	for(type = 0; type < 8; type++) {
+		if ( t == array_names[type][1] ) {
+			break;
+		}
+	}
+	if ( type > 8 )
+		goto out;
+
+	type += T_BOOLEAN;
+
+	r = allocTypeArray(type, size);
+
+out:
+	return r;
 }
 
 Object *allocObjectArray(Class *element_class, int length) {
@@ -2153,8 +2184,6 @@ Object *allocObjectArray(Class *element_class, int length) {
 }
 
 Object *allocTypeArray(int type, int size) {
-    static char *array_names[] = {"[Z", "[C", "[F", "[D", "[B",
-                                  "[S", "[I", "[J"};
     static int element_sizes[] = {1, 2, 4, 8, 1, 2, 4, 8};
     static Class *array_classes[8];
 
diff --git a/src/interp/engine/interp.c b/src/interp/engine/interp.c
index cfb6d67..f66af80 100644
--- a/src/interp/engine/interp.c
+++ b/src/interp/engine/interp.c
@@ -658,14 +658,15 @@ uintptr_t *executeJava() {
 #define ARRAY_LOAD_ARY *--ostack
 #endif
 
-#define ARRAY_LOAD(TYPE)                       \
-{                                              \
-    int idx = ARRAY_LOAD_IDX;                  \
-    Object *array = (Object *)ARRAY_LOAD_ARY;  \
-                                               \
-    NULL_POINTER_CHECK(array);                 \
-    ARRAY_BOUNDS_CHECK(array, idx);            \
-    PUSH_0(ARRAY_DATA(array, TYPE)[idx], 1);   \
+#define ARRAY_LOAD(TYPE)                         \
+{                                                \
+    int idx = ARRAY_LOAD_IDX;                    \
+    Object *array = (Object *)ARRAY_LOAD_ARY;    \
+                                                 \
+    NULL_POINTER_CHECK(array);                   \
+    NULL_POINTER_CHECK(ARRAY_DATA(array, TYPE)); \
+    ARRAY_BOUNDS_CHECK(array, idx);              \
+    PUSH_0(ARRAY_DATA(array, TYPE)[idx], 1);     \
 }
 
     DEF_OPC_012_2(
diff --git a/src/jam.h b/src/jam.h
index ce73df6..dbd003f 100644
--- a/src/jam.h
+++ b/src/jam.h
@@ -24,6 +24,7 @@
 #include <limits.h>
 #include <stdio.h>
 #include <time.h>
+#include <stddef.h>
 
 /* Configure options */
 #include "config.h"
@@ -420,6 +421,14 @@ typedef struct line_no_table_entry {
 
 typedef struct object Class;
 
+typedef struct array_object {
+   uintptr_t lock;
+   Class *class;
+   uintptr_t size;
+   uintptr_t data;
+   char contig_data[];
+} ArrayObject;
+
 typedef struct object {
    uintptr_t lock;
    Class *class;
@@ -784,7 +793,7 @@ typedef struct InitArgs {
 #define INST_DATA(obj, type, offset) *(type*)&((char*)obj)[offset]
 #define INST_BASE(obj, type)         ((type*)(obj+1))
 
-#define ARRAY_DATA(arrayRef, type)   ((type*)(((uintptr_t*)(arrayRef+1))+1)) 
+#define ARRAY_DATA(arrayRef, type)   (*((type**)(((uintptr_t*)(arrayRef+1))+1)))
 #define ARRAY_LEN(arrayRef)          *(uintptr_t*)(arrayRef+1)
 
 #define IS_CLASS(object)             (object->class && IS_CLASS_CLASS( \
@@ -910,6 +919,7 @@ extern int initialiseGC(InitArgs *args);
 extern Class *allocClass();
 extern Object *allocObject(Class *class);
 extern Object *allocTypeArray(int type, int size);
+extern Object *allocTypeArrayFromClassName(const char *className, int size);
 extern Object *allocObjectArray(Class *element_class, int size);
 extern Object *allocArray(Class *class, int size, int el_size);
 extern Object *allocMultiArray(Class *array_class, int dim, intptr_t *count);
-- 
1.9.1

[Jamvm-general] [PATCH] Additional methods from sun.misc.Unsafe

[<chr...@gm...>]

From: Christopher Friedt <ch...@mm...>

---
 src/classlib/gnuclasspath/natives.c |  2 ++
 src/jam.h                           |  4 ++++
 src/natives.c                       | 19 +++++++++++++++++++
 3 files changed, 25 insertions(+)

diff --git a/src/classlib/gnuclasspath/natives.c b/src/classlib/gnuclasspath/natives.c
index ca95e00..c52d590 100644
--- a/src/classlib/gnuclasspath/natives.c
+++ b/src/classlib/gnuclasspath/natives.c
@@ -1416,6 +1416,8 @@ VMMethod vm_stack_walker[] = {
 };
 
 VMMethod sun_misc_unsafe[] = {
+    {"addressSize",                 NULL, addressSize},
+    {"allocateInstance",            NULL, allocateInstance},
     {"objectFieldOffset",           NULL, objectFieldOffset},
     {"compareAndSwapInt",           NULL, compareAndSwapInt},
     {"compareAndSwapLong",          NULL, compareAndSwapLong},
diff --git a/src/jam.h b/src/jam.h
index ce73df6..d2a5af7 100644
--- a/src/jam.h
+++ b/src/jam.h
@@ -1235,6 +1235,10 @@ extern uintptr_t *putLong(Class *class, MethodBlock *mb, uintptr_t *ostack);
 extern uintptr_t *getLong(Class *class, MethodBlock *mb, uintptr_t *ostack);
 extern uintptr_t *putObject(Class *class, MethodBlock *mb, uintptr_t *ostack);
 
+extern uintptr_t *addressSize(Class *class, MethodBlock *mb,
+                              uintptr_t *ostack);
+extern uintptr_t *allocateInstance(Class *class, MethodBlock *mb,
+                                   uintptr_t *ostack);
 extern uintptr_t *objectFieldOffset(Class *class, MethodBlock *mb,
                                     uintptr_t *ostack);
 extern uintptr_t *compareAndSwapInt(Class *class, MethodBlock *mb,
diff --git a/src/natives.c b/src/natives.c
index 6afe705..782e6fd 100644
--- a/src/natives.c
+++ b/src/natives.c
@@ -110,6 +110,25 @@ void unlockSpinLock() {
     LOCKWORD_WRITE(&spinlock, 0);
 }
 
+uintptr_t *addressSize(Class *class, MethodBlock *mb, uintptr_t *ostack) {
+    *ostack++ = sizeof( uintptr_t );
+    return ostack;
+}
+
+uintptr_t *allocateInstance(Class *class, MethodBlock *mb, uintptr_t *ostack) {
+    class = (Class*)ostack[1];
+    ClassBlock *cb = CLASS_CB(class);
+    if(initClass(class) == NULL) {
+        return NULL;
+    }
+    Object *ob = allocTypeArrayFromClassName(cb->name, 0);
+    if ( NULL == ob ) {
+        ob = allocObject(class);
+    }
+    *ostack++ = ob;
+    return ostack;
+}
+
 uintptr_t *objectFieldOffset(Class *class, MethodBlock *mb, uintptr_t *ostack) {
     FieldBlock *fb = fbFromReflectObject((Object*)ostack[1]);
 
-- 
1.9.1

[Jamvm-general] [PATCH] Flexible Array Support

[<chr...@gm...>]

From: Christopher Friedt <chr...@gm...>

PATCH 1: Additional methods from sun.misc.Unsafe
PATCH 2: Array object modifications

This is a follow up from my RFC from some time ago:

http://goo.gl/FpTXUA

I've been using these results with quite a bit of success. My application is
memory mapping and painting the Linux framebuffer using only Java.
This method uses FileChannel.map().array() blows ByteBuffer operations out of
the water (since they make costly JNI calls each time). I've had speedups of
150x.

My first patchset unfortunately added a custom JNI call to facilitate
allocating what I call 'flexible' arrays (they are flexible to be contiguous
or discontiguous). Obviously custom JNI functions aren't portable. Therefore,
I converted the custom JNI to use sun.misc.Unsafe calls instead. Even though
this is 'unsafe', it seems to be slightly more portable than custom JNI.

I have complementary code for classpath which I would like to submit shortly
upstream, but I also wouldn't mind implementing this for OpenJDK compatibility
either.

C

Christopher Friedt (1):
  additional unsafe methods

 src/classlib/gnuclasspath/natives.c |  2 ++
 src/jam.h                           |  4 ++++
 src/natives.c                       | 19 +++++++++++++++++++
 3 files changed, 25 insertions(+)

-- 
1.9.1

Re: [Jamvm-general] Error building JAMVM on Mac

[Anitha B G. <ani...@gm...>]

On 26 October 2014 11:48, Robert Lougher <rob...@gm...> wrote:
> Yes it is active.  Unfortunately there's not a lot of help that I can
> give.  Macs used to be my main computers so they were well supported
> (under Linux and Mac OS X).  However, I no longer use Macs (my last
> Mac is from 2006 running OS X 10.5).  Luckily it appears that JamVM
> continued to be buildable on OS X at least until a year or two ago
> (judging by feedback) but obviously this is no longer the case.
> Without a current Mac there's not a lot I can do, so I'll probably
> remove support entirely from the next version.
>
> In case this is of help, I can see the following from your original post.
>
> 1) You no longer have gcc on Mac OS X.  The C compiler is actually
> Clang/LLVM, which is masquerading as gcc.  I have never built JamVM
> using LLVM as it uses several gcc features.  I no idea if JamVM
> compiles with LLVM.  Even if it does compile, it may not work.
>
> 2) By default, it is building 64-bit executables.  In the past, even
> on 64-bit machines, gcc on Mac OS X defaulted to 32-bit.  The error
> messages you are seeing is because JamVM by default (on Mac OS X)
> assumes it is being built for 32-bit.
>
> You could try configuring JamVM to use libffi, e.g. ./configure
> --enable-ffi.  As far as I'm aware, Mac OS X should still support
> libffi.  If this works you will have a 64-bit executable.
>
> Alternatively you could see if there is a flag to LLVM/Clang to force
> it to build in 32-bit mode.

Rob,

Finally, I am able to build 1.5.4 version forcing clang to build

[1] Using -m32.
[2] Removing "interp_cflags=-fno-reorder-blocks" (Clang does not like
this option)

Bare invocation worked so presumably everything should work though I
haven't run tests. As a side note --enable-libffi did not work either.
Configure (clang) complained that it does not know what the option
meant.

Thanks for your prompt help

-Anitha

[Jamvm-general] Implement JVM_FindClassFromCaller OpenJDK: 8015256: Better class accessibility

[Xerxes R. <xe...@gu...>]

The latest OpenJDK secuity update broke all alternative JVM support by
introducing a new symbol JVM_FindClassFromCaller

java -jamvm -version
java: relocation error:
/usr/lib/jvm/java-7-openjdk-amd64/jre/lib/amd64/libjava.so: symbol
JVM_FindClassFromCaller, version SUNWprivate_1.1 not defined in file
libjvm.so with link time reference


The OpenJDK source tree revealed the following information:

8015256: Better class accessibility

Summary: Improve protection domain check in forName()

http://hg.openjdk.java.net/jdk7u/jdk7u/hotspot/rev/61d1e75e0a58
http://hg.openjdk.java.net/jdk7u/jdk7u/jdk/rev/16cd2826a58f


JVM_FindClassFromCaller appears to work quite similar to
JVM_FindClassFromClassLoader with the twist that the protection domain that
belongs to the caller class argument shall be used during the lookup of the
class.

The attached patch implements JVM_FindClassFromCaller for JamVM with this
obvious TODO:
/* XXX The caller's protection domain should be used during   the
findClassFromClassLoader but there is no specification or   unit-test in
OpenJDK documenting the desired effect */

java -jamvm -version
java version "1.7.0_65"
OpenJDK Runtime Environment (IcedTea 2.5.3) (7u71-2.5.3-0ubuntu0.14.04.1)
JamVM (build 2.0.1-devel, inline-threaded interpreter)


Cheers
Xerxes

Re: [Jamvm-general] Error building JAMVM on Mac

[Robert L. <rob...@gm...>]

Yes it is active.  Unfortunately there's not a lot of help that I can
give.  Macs used to be my main computers so they were well supported
(under Linux and Mac OS X).  However, I no longer use Macs (my last
Mac is from 2006 running OS X 10.5).  Luckily it appears that JamVM
continued to be buildable on OS X at least until a year or two ago
(judging by feedback) but obviously this is no longer the case.
Without a current Mac there's not a lot I can do, so I'll probably
remove support entirely from the next version.

In case this is of help, I can see the following from your original post.

1) You no longer have gcc on Mac OS X.  The C compiler is actually
Clang/LLVM, which is masquerading as gcc.  I have never built JamVM
using LLVM as it uses several gcc features.  I no idea if JamVM
compiles with LLVM.  Even if it does compile, it may not work.

2) By default, it is building 64-bit executables.  In the past, even
on 64-bit machines, gcc on Mac OS X defaulted to 32-bit.  The error
messages you are seeing is because JamVM by default (on Mac OS X)
assumes it is being built for 32-bit.

You could try configuring JamVM to use libffi, e.g. ./configure
--enable-ffi.  As far as I'm aware, Mac OS X should still support
libffi.  If this works you will have a 64-bit executable.

Alternatively you could see if there is a flag to LLVM/Clang to force
it to build in 32-bit mode.

Rob.

On 26 October 2014 15:08, Anitha B Gollamudi <ani...@gm...> wrote:
> I am wondering if this mailing list is active...
>
> On 22 October 2014 12:53, Anitha B Gollamudi <ani...@gm...> wrote:
>> On 10 October 2014 05:00, Xerxes Rånby <xe...@za...> wrote:
>>>
>>> Den 2014-10-08 19:41, Anitha B Gollamudi skrev:
>>>>
>>>> Hi
>>>>
>>>> I took the latest jamvm and tried to make it on my macbook pro. I see
>>>> following errors. Looks like it is using x86_64 to build in i386. Any
>>>> idea what is going wrong? My specs are pasted at the end.
>>>>
>>>>
>>>> SEASs-MacBook-Pro:jamvm-1.5.4 anithagollamudi$ make
>>>
>>> Can you try JamVM 2.0.0?
>>> http://jamvm.sourceforge.net/
>>
>> This went to my spam and I just found it.
>>
>> 2.0.0 did not help either. It gave a configure error.
>>
>> SEASs-MacBook-Pro:jamvm-2.0.0 anithagollamudi$ ./configure
>> checking for a BSD-compatible install... /usr/bin/install -c
>> checking whether build environment is sane... yes
>> checking for a thread-safe mkdir -p... ./install-sh -c -d
>> checking for gawk... no
>> checking for mawk... no
>> checking for nawk... no
>> checking for awk... awk
>> checking whether make sets $(MAKE)... yes
>> checking build system type... x86_64-apple-darwin14.0.0
>> checking host system type... x86_64-apple-darwin14.0.0
>> configure: error: x86_64-apple-darwin14.0.0 not supported
>>
>>
>>>
>>> Cheers
>>> Xerxes
>>
>>
>>
>> --
>>  Anitha
>
>
>
> --
>  Anitha
>
> ------------------------------------------------------------------------------
> _______________________________________________
> Jamvm-general mailing list
> Jam...@li...
> https://lists.sourceforge.net/lists/listinfo/jamvm-general

Re: [Jamvm-general] Error building JAMVM on Mac

[Anitha B G. <ani...@gm...>]

I am wondering if this mailing list is active...

On 22 October 2014 12:53, Anitha B Gollamudi <ani...@gm...> wrote:
> On 10 October 2014 05:00, Xerxes Rånby <xe...@za...> wrote:
>>
>> Den 2014-10-08 19:41, Anitha B Gollamudi skrev:
>>>
>>> Hi
>>>
>>> I took the latest jamvm and tried to make it on my macbook pro. I see
>>> following errors. Looks like it is using x86_64 to build in i386. Any
>>> idea what is going wrong? My specs are pasted at the end.
>>>
>>>
>>> SEASs-MacBook-Pro:jamvm-1.5.4 anithagollamudi$ make
>>
>> Can you try JamVM 2.0.0?
>> http://jamvm.sourceforge.net/
>
> This went to my spam and I just found it.
>
> 2.0.0 did not help either. It gave a configure error.
>
> SEASs-MacBook-Pro:jamvm-2.0.0 anithagollamudi$ ./configure
> checking for a BSD-compatible install... /usr/bin/install -c
> checking whether build environment is sane... yes
> checking for a thread-safe mkdir -p... ./install-sh -c -d
> checking for gawk... no
> checking for mawk... no
> checking for nawk... no
> checking for awk... awk
> checking whether make sets $(MAKE)... yes
> checking build system type... x86_64-apple-darwin14.0.0
> checking host system type... x86_64-apple-darwin14.0.0
> configure: error: x86_64-apple-darwin14.0.0 not supported
>
>
>>
>> Cheers
>> Xerxes
>
>
>
> --
>  Anitha



-- 
 Anitha

Re: [Jamvm-general] Error building JAMVM on Mac

[Anitha B G. <ani...@gm...>]

On 10 October 2014 05:00, Xerxes Rånby <xe...@za...> wrote:
>
> Den 2014-10-08 19:41, Anitha B Gollamudi skrev:
>>
>> Hi
>>
>> I took the latest jamvm and tried to make it on my macbook pro. I see
>> following errors. Looks like it is using x86_64 to build in i386. Any
>> idea what is going wrong? My specs are pasted at the end.
>>
>>
>> SEASs-MacBook-Pro:jamvm-1.5.4 anithagollamudi$ make
>
> Can you try JamVM 2.0.0?
> http://jamvm.sourceforge.net/

This went to my spam and I just found it.

2.0.0 did not help either. It gave a configure error.

SEASs-MacBook-Pro:jamvm-2.0.0 anithagollamudi$ ./configure
checking for a BSD-compatible install... /usr/bin/install -c
checking whether build environment is sane... yes
checking for a thread-safe mkdir -p... ./install-sh -c -d
checking for gawk... no
checking for mawk... no
checking for nawk... no
checking for awk... awk
checking whether make sets $(MAKE)... yes
checking build system type... x86_64-apple-darwin14.0.0
checking host system type... x86_64-apple-darwin14.0.0
configure: error: x86_64-apple-darwin14.0.0 not supported


>
> Cheers
> Xerxes



-- 
 Anitha

Re: [Jamvm-general] RFC: JamVM 2.0.0, the first release of JamVM with support for OpenJDK: IcedTea 1, 2 & 3 update

[Andrew H. <gnu...@re...>]

----- Original Message -----
> "JamVM 2.0.0 is the first release of JamVM with support for OpenJDK.
> Although IcedTea already includes JamVM with OpenJDK support,
> this has been based on periodic snapshots of the development tree."
> http://sourceforge.net/projects/jamvm/files/jamvm/JamVM%202.0.0/README/view
> 
> Its clear that IcedTea should update to support this first JamVM release
> that now officially support OpenJDK!
> 
> I have attached patched to update IcedTea 1, 2 & 3 to use JamVM 2.0.0.
> 
> Unlike the JamVM GIT snapshots the jamvm-2.0.0.tar.gz release do not
> contain the autoconf.sh script,
> instead the release contain a pre-generated configure script.
> I have patched IcedTea to use the official JamVM releases from now on
> and thus only rely on the configure script to be present.
> 

Yes, I saw the release and thought of doing the same, but haven't had time.

I'm much happier with a proper release than hacky snapshots, so I hope this
will continue.

> 
> While testing I noted that icedtea 3 configure --enable-jamvm
> builds are currently broken,
> jdk 8 rev ce71b7e69937 added a new JVM requirement to implement
> JVM_GetTemporaryDirectory
> I have filed a bugreport to track this regression:
> http://icedtea.classpath.org/bugzilla/show_bug.cgi?id=2034
> 
> icedtea 3: configure --with-additional-vms=jamvm
> builds still succeed, libjava.so is then linked successfully against the
> hotspot libjvm.so
> openjdk.build/images/j2sdk-image$ ./bin/java -jamvm -version
> openjdk version "1.8.0_20"
> OpenJDK Runtime Environment (IcedTea 3.0.0pre02+rb27c53ea4272+) (Ubuntu
> build 1.8.0_20-b23)
> JamVM (build 2.0.0, inline-threaded interpreter)

Well, I believe the latter only avoids failing because it doesn't build the JDK against
the JamVM libjvm. This is why I don't support this way of building. It gives
a false hope that the build  has succeeded.

> 
> 
> icedtea 2: j2sdk-image$ ./bin/java -version
> java version "1.7.0_80"
> IcedTea Runtime Environment (2.6.0pre08+r17d957f63b11+) (Ubuntu build
> 1.7.0_80-b02)
> JamVM (build 2.0.0, inline-threaded interpreter)
> 
> 
> icedtea 1: j2sdk-image$ ./bin/java -version
> java version "1.6.0_32"
> IcedTea6 Runtime Environment (1.14.0pre+r3a715e42ffe4+) (Ubuntu build
> 1.6.0_32-b32)
> JamVM (build 2.0.0, inline-threaded interpreter)
> 
> 
> JamVM upstream and GIT its located at sourceforge.
> http://sourceforge.net/p/jamvm
> http://sourceforge.net/p/jamvm/code/ci/master/tree/
> http://sourceforge.net/p/jamvm/code/ci/master/tree/INSTALL
> 

There's an issue due to a new function introduced by the latest security
update:

http://icedtea.classpath.org/bugzilla/show_bug.cgi?id=2032

(that's for CACAO, you probably want some for JamVM too)

so it's probably best to wait until this is resolved to save on having to
do two updates.

> Cheers
> Xerxes
> 
> 
> 
> 

-- 
Andrew :)

Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

PGP Key: 248BDC07 (https://keys.indymedia.org/)
Fingerprint = EC5A 1F5E C0AD 1D15 8F1F  8F91 3B96 A578 248B DC07

[Jamvm-general] RFC: JamVM 2.0.0, the first release of JamVM with support for OpenJDK: IcedTea 1, 2 & 3 update

[Xerxes R. <xe...@za...>]

"JamVM 2.0.0 is the first release of JamVM with support for OpenJDK.
Although IcedTea already includes JamVM with OpenJDK support,
this has been based on periodic snapshots of the development tree."
http://sourceforge.net/projects/jamvm/files/jamvm/JamVM%202.0.0/README/view

Its clear that IcedTea should update to support this first JamVM release 
that now officially support OpenJDK!

I have attached patched to update IcedTea 1, 2 & 3 to use JamVM 2.0.0.

Unlike the JamVM GIT snapshots the jamvm-2.0.0.tar.gz release do not 
contain the autoconf.sh script,
instead the release contain a pre-generated configure script.
I have patched IcedTea to use the official JamVM releases from now on 
and thus only rely on the configure script to be present.


While testing I noted that icedtea 3 configure --enable-jamvm
builds are currently broken,
jdk 8 rev ce71b7e69937 added a new JVM requirement to implement
JVM_GetTemporaryDirectory
I have filed a bugreport to track this regression:
http://icedtea.classpath.org/bugzilla/show_bug.cgi?id=2034

icedtea 3: configure --with-additional-vms=jamvm
builds still succeed, libjava.so is then linked successfully against the 
hotspot libjvm.so
openjdk.build/images/j2sdk-image$ ./bin/java -jamvm -version
openjdk version "1.8.0_20"
OpenJDK Runtime Environment (IcedTea 3.0.0pre02+rb27c53ea4272+) (Ubuntu 
build 1.8.0_20-b23)
JamVM (build 2.0.0, inline-threaded interpreter)


icedtea 2: j2sdk-image$ ./bin/java -version
java version "1.7.0_80"
IcedTea Runtime Environment (2.6.0pre08+r17d957f63b11+) (Ubuntu build 
1.7.0_80-b02)
JamVM (build 2.0.0, inline-threaded interpreter)


icedtea 1: j2sdk-image$ ./bin/java -version
java version "1.6.0_32"
IcedTea6 Runtime Environment (1.14.0pre+r3a715e42ffe4+) (Ubuntu build 
1.6.0_32-b32)
JamVM (build 2.0.0, inline-threaded interpreter)


JamVM upstream and GIT its located at sourceforge.
http://sourceforge.net/p/jamvm
http://sourceforge.net/p/jamvm/code/ci/master/tree/
http://sourceforge.net/p/jamvm/code/ci/master/tree/INSTALL

Cheers
Xerxes

Re: [Jamvm-general] Error building JAMVM on Mac

[Xerxes R. <xe...@za...>]

Den 2014-10-08 19:41, Anitha B Gollamudi skrev:
> Hi
>
> I took the latest jamvm and tried to make it on my macbook pro. I see
> following errors. Looks like it is using x86_64 to build in i386. Any
> idea what is going wrong? My specs are pasted at the end.
>
>
> SEASs-MacBook-Pro:jamvm-1.5.4 anithagollamudi$ make
Can you try JamVM 2.0.0?
http://jamvm.sourceforge.net/

Cheers
Xerxes

[Jamvm-general] Error building JAMVM on Mac

[Anitha B G. <ani...@gm...>]

Hi

I took the latest jamvm and tried to make it on my macbook pro. I see
following errors. Looks like it is using x86_64 to build in i386. Any
idea what is going wrong? My specs are pasted at the end.


SEASs-MacBook-Pro:jamvm-1.5.4 anithagollamudi$ make
Making all in src
/Library/Developer/CommandLineTools/usr/bin/make  all-recursive
Making all in os
Making all in darwin
Making all in i386
/bin/sh ../../../../libtool --tag=CC   --mode=compile gcc
-DHAVE_CONFIG_H -I. -I../../../../src  -I../../../../src   -g -O2 -MT
dll_md.lo -MD -MP -MF .deps/dll_md.Tpo -c -o dll_md.lo dll_md.c
 gcc -DHAVE_CONFIG_H -I. -I../../../../src -I../../../../src -g -O2
-MT dll_md.lo -MD -MP -MF .deps/dll_md.Tpo -c dll_md.c  -fno-common
-DPIC -o .libs/dll_md.o
dll_md.c:70:19: error: invalid operand for instruction
    asm volatile ("subl %0,%%esp" :: "r" ((aligned - tot_args) * sizeof(u4))
                  ^
<inline asm>:1:7: note: instantiated into assembly here
        subl %rdx,%esp
             ^~~~~
dll_md.c:74:23: error: instruction requires: Not 64-bit mode
        asm volatile ("pushl %0" :: "m" (*--opntr) : "sp");
                      ^
<inline asm>:1:2: note: instantiated into assembly here
        pushl (%rdx)

...............

$ gcc -v
Configured with: --prefix=/Library/Developer/CommandLineTools/usr
--with-gxx-include-dir=/usr/include/c++/4.2.1
Apple LLVM version 6.0 (clang-600.0.51) (based on LLVM 3.5svn)
Target: x86_64-apple-darwin13.4.0
Thread model: posix

..........

kern.version: Darwin Kernel Version 13.4.0: Sun Aug 17 19:50:11 PDT
2014; root:xnu-2422.115.4~1/RELEASE_X86_64

machdep.cpu.brand_string: Intel(R) Core(TM) i7-4980HQ CPU @ 2.80GHz

--
 Anitha


-- 
 Anitha

Re: [Jamvm-general] What's the status of jamvm 1.6.0?

[Andrew H. <gnu...@re...>]

----- Original Message -----
> Hi Thomas,
> 
> On 19 August 2014 15:29, Thomas De Schampheleire
> <pat...@gm...> wrote:
> > Hi Rob,
> >
> > I noticed that JamVM 2.0.0 has been released now, thanks for that!
> >
> > Some observations though:
> >
> > - the Sourceforge web site http://jamvm.sourceforge.net/, the summary
> > page http://sourceforge.net/projects/jamvm/ and the Download page
> > http://sourceforge.net/projects/jamvm/files/ all still refer to 1.5.4
> > as the latest release. I think this should be updated too.
> >
> 
> Website updated.  I can't update the Download page - I added a new
> release, and it should point to it as the latest automatically.  I'll
> raise an issue with SourceForge.
> 
> > - I have not seen any announcement on the JamVM mailing list. Is it no
> > longer in use? Googling explicitly, I found only an announcement on
> > your blog http://draenog.blogspot.be/2014/08/jamvm-200-released.html.
> >
> 
> The mailing list is still in use.  However, judging by page visit
> counts, the blog has much greater coverage (links from Planet JDK,
> Planet Classpath, and various other developer sites).
> 
> Rob.
> 
> > Thanks,
> > Thomas
> 
> ------------------------------------------------------------------------------
> _______________________________________________
> Jamvm-general mailing list
> Jam...@li...
> https://lists.sourceforge.net/lists/listinfo/jamvm-general
> 

Great news! I'll rebase the various IcedTea trees to JamVM 2.0.0.
-- 
Andrew :)

Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

PGP Key: 248BDC07 (https://keys.indymedia.org/)
Fingerprint = EC5A 1F5E C0AD 1D15 8F1F  8F91 3B96 A578 248B DC07

Re: [Jamvm-general] What's the status of jamvm 1.6.0?

[Guillermo R. <gro...@in...>]

Hi Robert,

This is excellent news. 
The benefits for OpenJDK users are obvious of course. Can you summarize what we can expect in terms of stability, bugfixes, testing coverage, etc. when using GNU Classpath (relative to 1.5.4)? This was not clear to me from the release notes. 

Thank you,

Guillermo Rodriguez
Enviado desde mi iPhone

El 19/08/2014, a las 18:32, Robert Lougher <rob...@gm...> escribió:

> Hi Thomas,
> 
> On 19 August 2014 15:29, Thomas De Schampheleire
> <pat...@gm...> wrote:
>> Hi Rob,
>> 
>> I noticed that JamVM 2.0.0 has been released now, thanks for that!
>> 
>> Some observations though:
>> 
>> - the Sourceforge web site http://jamvm.sourceforge.net/, the summary
>> page http://sourceforge.net/projects/jamvm/ and the Download page
>> http://sourceforge.net/projects/jamvm/files/ all still refer to 1.5.4
>> as the latest release. I think this should be updated too.
>> 
> 
> Website updated.  I can't update the Download page - I added a new
> release, and it should point to it as the latest automatically.  I'll
> raise an issue with SourceForge.
> 
>> - I have not seen any announcement on the JamVM mailing list. Is it no
>> longer in use? Googling explicitly, I found only an announcement on
>> your blog http://draenog.blogspot.be/2014/08/jamvm-200-released.html.
>> 
> 
> The mailing list is still in use.  However, judging by page visit
> counts, the blog has much greater coverage (links from Planet JDK,
> Planet Classpath, and various other developer sites).
> 
> Rob.
> 
>> Thanks,
>> Thomas
> 
> ------------------------------------------------------------------------------
> _______________________________________________
> Jamvm-general mailing list
> Jam...@li...
> https://lists.sourceforge.net/lists/listinfo/jamvm-general

Re: [Jamvm-general] What's the status of jamvm 1.6.0?

[Robert L. <rob...@gm...>]

Hi Thomas,

On 19 August 2014 15:29, Thomas De Schampheleire
<pat...@gm...> wrote:
> Hi Rob,
>
> I noticed that JamVM 2.0.0 has been released now, thanks for that!
>
> Some observations though:
>
> - the Sourceforge web site http://jamvm.sourceforge.net/, the summary
> page http://sourceforge.net/projects/jamvm/ and the Download page
> http://sourceforge.net/projects/jamvm/files/ all still refer to 1.5.4
> as the latest release. I think this should be updated too.
>

Website updated.  I can't update the Download page - I added a new
release, and it should point to it as the latest automatically.  I'll
raise an issue with SourceForge.

> - I have not seen any announcement on the JamVM mailing list. Is it no
> longer in use? Googling explicitly, I found only an announcement on
> your blog http://draenog.blogspot.be/2014/08/jamvm-200-released.html.
>

The mailing list is still in use.  However, judging by page visit
counts, the blog has much greater coverage (links from Planet JDK,
Planet Classpath, and various other developer sites).

Rob.

> Thanks,
> Thomas

Re: [Jamvm-general] What's the status of jamvm 1.6.0?

[Thomas De S. <pat...@gm...>]

Hi Rob,

On Wed, Jun 11, 2014 at 11:35 AM, Robert Lougher <rob...@gm...> wrote:
> Hi Thomas,
>
> Code hosting has moved back to Sourceforge, details here:
> http://sourceforge.net/p/jamvm/code/ci/master/tree/
>
> I found an issue with nashorn while testing OpenJDK 8 which required
> rewriting some of the invokedynamic code.  This is now done, and I'm
> working on the release notes.

I noticed that JamVM 2.0.0 has been released now, thanks for that!

Some observations though:

- the Sourceforge web site http://jamvm.sourceforge.net/, the summary
page http://sourceforge.net/projects/jamvm/ and the Download page
http://sourceforge.net/projects/jamvm/files/ all still refer to 1.5.4
as the latest release. I think this should be updated too.

- I have not seen any announcement on the JamVM mailing list. Is it no
longer in use? Googling explicitly, I found only an announcement on
your blog http://draenog.blogspot.be/2014/08/jamvm-200-released.html.

Thanks,
Thomas

Re: [Jamvm-general] JamVM /JDWP

[Christopher F. <chr...@gm...>]

Incidentally, it has been rumored for some time, but it would seem that
Dalvik is now being deprecated by Google on the Android platform.

http://www.xda-developers.com/android/breaking-next-major-version-of-android-to-finally-remove-dalvik-and-set-art-as-default/

I wonder if it would make sense for e.g. a GSOC student to potentially work
on introducing optimizations in the Dalvik VM back into JamVM / Classpath,
e.g. a JIT - at least conceptually similar algorithms in any case, as the
codebase has changed dramatically since it was "forked" originally.

©

Sent from my Android

[Jamvm-general] JamVM /JDWP

[Christopher F. <chr...@gm...>]

Hi list

I'd like to propose that a bunch of us hack together support for the Java
Debug Wire Protocol[1] in JamVM.

I believe GNU Class path already has done a fair amount of work on this,
but I could be mistaken.

JDWP would be a handy feature to --enable or --disable at configure time
for JamVM, since adding debugging capabilities would potentially introduce
additional size and speed concerns. However, it could be very useful for
developers during coding and debug phases of a project's lifecycle.
Additionally, due to JDWP flexibility, it would enable people to remote
debug over a socket and step-debug code in e.g. Eclipse.

I will need this feature for at least one open-source project I'm working
on [2][3], but I feel as though the rest of the community could benefit
from it as well, if their projects are tied to JamVM / Classpath.

Incidentally, is anyone aware as to whether JDWP is required for profiling
a JVM?

Cheers,

©

Sent from my Android

[1] http://docs.oracle.com/javase/1.5.0/docs/guide/jpda/jdwp-spec.html
[2] https://plus.google.com/107058277812335526223/posts/N4eVKvn2oL4
[3] https://github.com/cfriedt/fb4j

Re: [Jamvm-general] fb4j

[Christopher F. <chr...@gm...>]

Here is my post from G+

https://plus.google.com/107058277812335526223/posts/N4eVKvn2oL4

©

Sent from my Android
On Dec 31, 2013 2:41 PM, "Christopher Friedt" <chr...@gm...> wrote:

> Hi list,
>
> Shameless self-promotion that might be of interest for embedded Java
> developers.
>
> https://github.com/cfriedt/fb4j
>
> It wouldn't be possible without JamVM of course ;-)
>
> Happy New Year!
>
> C
>

Re: [Jamvm-general] What's the status of jamvm 1.6.0?

[Robert L. <rob...@gm...>]

Hi Thomas,

Code hosting has moved back to Sourceforge, details here:
http://sourceforge.net/p/jamvm/code/ci/master/tree/

I found an issue with nashorn while testing OpenJDK 8 which required
rewriting some of the invokedynamic code.  This is now done, and I'm
working on the release notes.

Rob.

On 11 June 2014 08:34, Thomas De Schampheleire
<pat...@gm...> wrote:
> Hi Robert,
>
> On Fri, May 16, 2014 at 1:23 AM, Robert Lougher <rob...@gm...> wrote:
>> On 8 May 2014 12:39, Thomas De Schampheleire <pat...@gm...> wrote:
>>> Hi Robert,
>>>
>>> On Thu, May 8, 2014 at 1:20 PM, Robert Lougher <rob...@gm...> wrote:
>>>>My intention was to do this over the weekend, so hopefully
>>>> if no problems show up there should be a new release in a couple of
>>>> days.
>>>>
>>>
>>> That sounds great, thanks a lot! This will definitely help the
>>> Buildroot project.
>>>
>>
>> Quick update.  Testing done, no regressions found compared to 1.5.4
>> (with GNU Classpath 0.99).  Should be on track to do the release this
>> weekend.
>>
>
>
> Any further news?
>
> I wanted to check the git at Berlios, but seems they finally shut down.
> Where is the jamvm git repo located now?
>
> Thanks,
> Thomas

Re: [Jamvm-general] What's the status of jamvm 1.6.0?

[Thomas De S. <pat...@gm...>]

Hi Robert,

On Fri, May 16, 2014 at 1:23 AM, Robert Lougher <rob...@gm...> wrote:
> On 8 May 2014 12:39, Thomas De Schampheleire <pat...@gm...> wrote:
>> Hi Robert,
>>
>> On Thu, May 8, 2014 at 1:20 PM, Robert Lougher <rob...@gm...> wrote:
>>>My intention was to do this over the weekend, so hopefully
>>> if no problems show up there should be a new release in a couple of
>>> days.
>>>
>>
>> That sounds great, thanks a lot! This will definitely help the
>> Buildroot project.
>>
>
> Quick update.  Testing done, no regressions found compared to 1.5.4
> (with GNU Classpath 0.99).  Should be on track to do the release this
> weekend.
>


Any further news?

I wanted to check the git at Berlios, but seems they finally shut down.
Where is the jamvm git repo located now?

Thanks,
Thomas

Re: [Jamvm-general] What's the status of jamvm 1.6.0?

[Andïï <gnu...@me...>]

On 22 May 2014 00:48, Ulrich Herberg <ul...@he...> wrote:
> On Wed, May 21, 2014 at 4:45 PM, Andïï <gnu...@me...> wrote:
>> Grab the latest IcedTea release:
>>
>> http://blog.fuseyism.com/index.php/2014/04/16/security-icedtea-2-4-7-for-openjdk-7-released/
>>
>> and build with ./configure --enable-jamvm
>
>
> Thanks. Since I use buildroot with a cross-compiler for ARM, I was
> wondering if that is easily possible. I tried setting up Icedtea
> inside buildroot (not sure if that is the right approach), but that is
> quite complicated (many dependencies).
>
> Ulrich

Maybe I can say more if you describe your setup in more detail. I don't
know what 'buildroot' is.

OpenJDK isn't great at cross-compiling, though I did manage to get an x86 build
on x86_64 to work today. So it should be possible, in theory.
-- 
Andii :-)

Re: [Jamvm-general] What's the status of jamvm 1.6.0?

[Ulrich H. <ul...@he...>]

On Wed, May 21, 2014 at 4:45 PM, Andïï <gnu...@me...> wrote:
> Grab the latest IcedTea release:
>
> http://blog.fuseyism.com/index.php/2014/04/16/security-icedtea-2-4-7-for-openjdk-7-released/
>
> and build with ./configure --enable-jamvm


Thanks. Since I use buildroot with a cross-compiler for ARM, I was
wondering if that is easily possible. I tried setting up Icedtea
inside buildroot (not sure if that is the right approach), but that is
quite complicated (many dependencies).

Ulrich