patch to bring java vtables closer to g++ abi conformance

Per Bothner per@bothner.com
Wed Jan 23 17:56:00 GMT 2002


Richard Henderson wrote:
> On Wed, Jan 23, 2002 at 01:11:39PM -0800, Per Bothner wrote:
> 
>>I agree, that should probably be null_pointer_node, though it
>>doesn't really matter..  (It gets cast to ptr_type_node anyway.)
>>
> 
> Where?  I don't see that.

Well, there is code in output_constructor that figures out the
fieldsize.  However, the code is convoluted, and it may depend on
setting the TREE_PURPOSE, so counting on it seems fragile.

> And it _does_ matter.  The types of the data in static initializers
> must be correct, lest varasm.c get confused.  So if you use
> integer_zero_node when you really meant a long, the wrong amount
> of data will be emitted for 64-bit targets.

Ok - I changed it to use null_pointer_node, which should be fine,
except for systems where the null pointer isn't all zeros - which
I'm sure will break in other ways.
-- 
	--Per Bothner
per@bothner.com   http://www.bothner.com/per/



More information about the Java mailing list