Compatibility relation for "native int"

Thread view:

Enable EMail Alerts

Start New Thread

Thread rating:

George - 06 Nov 2005 23:59 GMT

I'm interested in the compatibility relation (used by the bytecode verifier)
which involves the type "native int".

So, is int32 compatible with native int? i.e. is int32 a subtype of native
int?

The few examples I run show that the bytecode verifier accepts an "int32"
where a "native int" is expected.

The same question also for int64: is int64 a subtype of native int?

Many thanks for any answer.

Reply to this Message

Richard Grimes - 07 Nov 2005 13:57 GMT

> I'm interested in the compatibility relation (used by the bytecode
> verifier) which involves the type "native int".
[quoted text clipped - 8 lines]
>
> Many thanks for any answer.

The ECMA spec says 'these types exist as CLI types, but the CLI maps
each to the native size for a specific processor' (Partitian 1, 12.1.1).
So although native int is the same as int32 on x86, it may not be the
same on another priocessor. It's like the difference between int and
__int32 in C++

Richard

Signature

http://www.grimes.demon.co.uk/workshops/fusionWS.htm
http://www.grimes.demon.co.uk/workshops/securityWS.htm

Reply to this Message

George - 07 Nov 2005 14:50 GMT

> The ECMA spec says 'these types exist as CLI types, but the CLI maps
> each to the native size for a specific processor' (Partitian 1, 12.1.1).

I'm aware of this.

It's not clear to me why the bytecode verifier accepts int32 or int64 where
an native int is expected. Does this mean that the verifier considers int32
and int64 compatible with native int?

Reply to this Message

Richard Grimes - 07 Nov 2005 16:23 GMT

>> The ECMA spec says 'these types exist as CLI types, but the CLI maps
>> each to the native size for a specific processor' (Partitian 1,
[quoted text clipped - 4 lines]
> It's not clear to me why the bytecode verifier accepts int32 or int64
> where an native int is expected.

Well, according to the spec int32 should be accepted on a processor
where ints are 32-bits and in64 should be accepted on a processor where
ints are 64-bits. Are you saying that on x86 the verifier will accept
int32 or int64? If so, then that is a serious bug.

> Does this mean that the verifier
> considers int32 and int64 compatible with native int?

Only on the respective processors. It shouldn't accept either on the
same processor. Do you have examples of what you are talking about?

Richard

Signature

http://www.grimes.demon.co.uk/workshops/fusionWS.htm
http://www.grimes.demon.co.uk/workshops/securityWS.htm

Reply to this Message

George - 07 Nov 2005 21:28 GMT

It's not clear to me that the verification is arhitecture dependent ...

> Do you have examples of what you are talking about?

yes,

ECMA, Partition III, chapter 3.18: "The brtrue instruction transfers control
to target if value (of type native int) is nonzero (true)." However, the
verifier accepts also values of types "int32" and "int64".

both fragments are accepted by the verifier:

ldc.i4 2
brtrue target

ldc.i8 2
brtrue target

Reply to this Message

Compatibility relation for "native int"

Free Magazines