Dalibor Topic wrote:
On 9/1/11 1:11 PM, Joe Darcy wrote:
Hello.
I hereby request approval to backport a trio of signature parsing fixes from
JDK 8 to 7 update:
6476261: (reflect) GenericSignatureFormatError When signature includes
nested inner classes
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6476261
6832374: (reflect) malformed signature can cause parser to go into infinite
loop
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=6832374
7052898: (reflect) SignatureParser will accept strings outside of the grammar
http://bugs.sun.com/bugdatabase/view_bug.do?bug_id=7052898
Unfortunately, the last bug comes up as not available for me. Could you
describe it in your own words,
and let us know whether this is a regression, and what its priority is?
Sure; the legal syntax for a throws clause is a proper subset of the
full FieldTypeSignature production. The old code did not make that
distinction, improperly allowing arrays in the throws section. The new
code fixes this issue by adding a b0olean argument in
parseFieldTypeSignature(boolean allowArrays). The problem is not a
regression, but is is included as a unit with the other two fixes.
Webrev:
http://cr.openjdk.java.net/~darcy/6476261.0/
The code for 7 update is trivially different than the code for JDK 8 since the
JDK 8 code takes advantage of a new constructor added to
GenericSignatureFormatError; here is the diff to SignatureParser.java:
123,126c123,124
< if (DEBUG)
< System.out.println("Signature Parse error: " + errorMsg +
< "\n\tRemaining input: " + remainder());
< return new GenericSignatureFormatError();
---
return new GenericSignatureFormatError("Signature Parse error: " +
errorMsg +
"\n\tRemaining input: " +
remainder());
Thanks - I am a bit puzzled since the new jdk 8 constructor seems to still be used in the
posted webrev.
Sorry; I've posted the 7 update version of the webrev at:
http://cr.openjdk.java.net/~darcy/6476261.7u/
-Joe
