Serguei,
Line 91, 92 are same, when used after here, they are cast to short and
byte according format "JJ" or "j".
for 78 and 79, we need to get accurate index, so need to mask
corresponding upper bytes since we do not need to check format . If you
check usage after line number 91,92, you will find they are treated same
as 78,79.
I will add _invokehandle since it got missed after permgen removal,
Coleen pointed out this.
Will send webrev tomorrow.
Thanks
Yumin
On 10/17/2012 10:06 PM, [email protected] wrote:
Hi Yumin,
I've got another question related to masking.
No consistency between methods getConstantPoolIndexFromRefMap and
getConstantPoolIndex.
Compare the lines 78+79 with 91+92:
74 protected short getConstantPoolIndexFromRefMap(int rawcode, int bci) {
75 int refIndex;
76 String fmt = Bytecodes.format(rawcode);
77 switch (fmt.length()) {
78 case 2: refIndex =0xFF& method.getBytecodeByteArg(bci);
break;
79 case 3: refIndex =0xFFFF&
bytes.swapShort(method.getBytecodeShortArg(bci)); break;
. . .
86 protected short getConstantPoolIndex(int rawcode, int bci) {
87 // get ConstantPool index from ConstantPoolCacheIndex at given bci
88 String fmt = Bytecodes.format(rawcode);
89 int cpCacheIndex;
90 switch (fmt.length()) {
91 case 2: cpCacheIndex = method.getBytecodeByteArg(bci); break;
92 case 3: cpCacheIndex = method.getBytecodeShortArg(bci); break;
The refIndex is positive so that there is no point to mask byte and short
values at lines 78,79.
Also one more question:
Why these functions are protected, not private?
The rest looks good.
Thanks,
Serguei
On 10/17/12 5:25 PM, Yumin Qi wrote:
Hi,
As suggested by David, I created a new bug
8001055: Bytes.swap should follow big endian
to address the issued caused by 6879063 push.
new webrevs:
http://cr.openjdk.java.net/~minqi/8000818
http://cr.openjdk.java.net/~minqi/8001055
Thanks
Yumin
On 10/16/2012 9:51 PM, Yumin Qi wrote:
Hi, all
May I have your codereview on
http://cr.openjdk.java.net/~minqi/8000818/
<http://cr.openjdk.java.net/%7Eminqi/8000818/>
8000818: SA constant pool need to reference to reference map after
permgen removal
Summary: After permgen removal, constant pool changed to put _ldc
and _ldc_w (fast_ldc and fast_ldcw) index to reference map, no
longer calculated via constant pool cache.
Also, there is a mistake in 6879063: SA should use hsdis. Bytes.swap
should only check if the underlying platform is big endian since
java code follows big endian. Revert it back to its orginal form,
else it will fail ClassDump.
Reviewed-by:
Contributed-by: [email protected]
Thanks
Yumin
