Allow 1.7 class file format

Allow 1.7 class file format

[Andrew Haley]

There's no real reason to disallow 1.7 class files.  Anyone trying to
use any of the new opcodes or new constants will fail verification,
and most Java programs don't use them.

Andrew.


012-04-23  Andrew Haley  <aph@redhat.com>

	* defineclass.cc (MAJOR_1_7,  MINOR_1_7): New.
	(parse): Allow MAJOR_1_7 classfile version.

Index: defineclass.cc
===================================================================
--- defineclass.cc	(revision 186103)
+++ defineclass.cc	(working copy)
@@ -361,6 +361,8 @@
 #define MINOR_1_5  0
 #define MAJOR_1_6 50
 #define MINOR_1_6  0
+#define MAJOR_1_7 51
+#define MINOR_1_7  0

 void
 _Jv_ClassReader::parse ()
@@ -371,8 +373,8 @@

   int minor_version = read2u ();
   int major_version = read2u ();
-  if (major_version < MAJOR_1_1 || major_version > MAJOR_1_6
-      || (major_version == MAJOR_1_6 && minor_version > MINOR_1_6))
+  if (major_version < MAJOR_1_1 || major_version > MAJOR_1_7
+      || (major_version == MAJOR_1_7 && minor_version > MINOR_1_7))
     throw_class_format_error ("unrecognized class file version");
   is_15 = (major_version >= MAJOR_1_5);

Re: Allow 1.7 class file format

[Andrew Hughes]

----- Original Message -----
> There's no real reason to disallow 1.7 class files.  Anyone trying to
> use any of the new opcodes or new constants will fail verification,
> and most Java programs don't use them.
> 

Seems very sensible to me.  I was thinking we'd need something like this for
bootstrapping if nothing else.

FWIW, I think gcj is the only Classpath VM that checks the range of this at all
and it's always seemed fairly incompatible with a FOSS model of working to me
i.e. if there's a real issue with not being able to read a class, it's easier
to dig into the code and debug it when it hits the unknown bytecode
rather than on some arbitrary version check.  As you say, in the majority of cases, 1.7
versioned bytecode will only contain bytecodes from <=1.6 anyway.

> Andrew.
> 
> 
> 012-04-23  Andrew Haley  <aph@redhat.com>
> 
> 	* defineclass.cc (MAJOR_1_7,  MINOR_1_7): New.
> 	(parse): Allow MAJOR_1_7 classfile version.
> 
> Index: defineclass.cc
> ===================================================================
> --- defineclass.cc	(revision 186103)
> +++ defineclass.cc	(working copy)
> @@ -361,6 +361,8 @@
>  #define MINOR_1_5  0
>  #define MAJOR_1_6 50
>  #define MINOR_1_6  0
> +#define MAJOR_1_7 51
> +#define MINOR_1_7  0
> 
>  void
>  _Jv_ClassReader::parse ()
> @@ -371,8 +373,8 @@
> 
>    int minor_version = read2u ();
>    int major_version = read2u ();
> -  if (major_version < MAJOR_1_1 || major_version > MAJOR_1_6
> -      || (major_version == MAJOR_1_6 && minor_version > MINOR_1_6))
> +  if (major_version < MAJOR_1_1 || major_version > MAJOR_1_7
> +      || (major_version == MAJOR_1_7 && minor_version > MINOR_1_7))
>      throw_class_format_error ("unrecognized class file version");
>    is_15 = (major_version >= MAJOR_1_5);
> 
> 
> 

-- 
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: Allow 1.7 class file format

[Andrew Haley]

On 04/23/2012 05:58 PM, Andrew Hughes wrote:
> FWIW, I think gcj is the only Classpath VM that checks the range of this at all
> and it's always seemed fairly incompatible with a FOSS model of working to me
> i.e. if there's a real issue with not being able to read a class, it's easier
> to dig into the code and debug it when it hits the unknown bytecode
> rather than on some arbitrary version check.

I don't know why that check is there.  I suppose it's always been
easier to update it rather than just delete the check.

Andrew.

Re: Allow 1.7 class file format

[Andrew Haley]

On 04/23/2012 02:30 PM, Andrew Haley wrote:
> 012-04-23  Andrew Haley  <aph@redhat.com>
> 
> 	* defineclass.cc (MAJOR_1_7,  MINOR_1_7): New.
> 	(parse): Allow MAJOR_1_7 classfile version.

Committed to 4.7 release branch.

Andrew.