[Bug 100017] XML encoder can cause a StackOverflowError

[Joe Darcy]

Hello.

Andrew John Hughes wrote:
> On 09:50 Fri 15 Jan     , Joe Wang wrote:
>> Thanks Alan. Yes, I did receive the bugzilla notice.
>>
>> Kelly told me that he applied the same jaxp source tarball to OpenJDK6. 
>> Kelly, could you tell Andrew how to include the jaxp source tarball when 
>> building OpenJDK6?
>>
> 
> The tarball jdk6-jaxp-2009_10_27.zip used by the OpenJDK6 build does not
> include the fix.  We are still having to apply the patch:
> 
> diff -Nru openjdk.orig/jaxp/build.properties openjdk/jaxp/build.properties
> --- openjdk.orig/jaxp/build.properties	     2009-12-08 17:42:33.000000000 +0000
> +++ openjdk/jaxp/build.properties	     2009-12-08 17:43:03.000000000 +0000
> @@ -73,6 +73,9 @@
>  # Where patches to drop bundle sources live
>  patches.dir=patches
>  
> +# Patches to apply
> +jaxp_src.patch.list=xml-encodinginfo.patch
> +
>  # Sanity information
>  sanity.info= Sanity Settings:${line.separator}\
>    ant.home=${ant.home}${line.separator}\
> diff -Nru openjdk.orig/jaxp/patches/jaxp_src/xml-encodinginfo.patch openjdk/jaxp/patches/jaxp_src/xml-encodinginfo.patch
> --- openjdk.orig/jaxp/patches/jaxp_src/xml-encodinginfo.patch	    1970-01-01 01:00:00.000000000 +0100
> +++ openjdk/jaxp/patches/jaxp_src/xml-encodinginfo.patch	    2009-12-08 17:41:58.000000000 +0000
> @@ -0,0 +1,18 @@
> +diff -Nru src/com/sun/org/apache/xml/internal/serializer/EncodingInfo.java src.new/com/sun/org/apache/xml/internal/serializer/EncodingInfo.java
> +--- src/com/sun/org/apache/xml/internal/serializer/EncodingInfo.java	    2009-10-27 21:54:16.000000000 +0000
> ++++ src.new/com/sun/org/apache/xml/internal/serializer/EncodingInfo.java   2009-12-08 17:40:14.000000000 +0000
> +@@ -326,9 +326,11 @@
> +             m_last = last;
> + 
> +             // Set the range of unicode values that this object
> +-            // explicitly manages
> +-            m_explFirst = codePoint;
> +-            m_explLast = codePoint + (RANGE-1);
> ++            // explicitly manages. Align the explicitly managed values
> ++            // to RANGE so multiple EncodingImpl objects dont manage the same 
> ++            // values.
> ++            m_explFirst = codePoint / RANGE * RANGE;
> ++            m_explLast = m_explFirst + (RANGE-1);
> + 
> +             m_encoding = encoding;
> + 

Joe (Wang), I'll let you decide on the appropriateness of the fix for 
jaxp.  If the fix is good, a new jaxp bundle for OpenJDK 6 would be fine 
by me.

> BTW, we have a testcase for this.  Would it be possible to add this to
> the JDK tree?  The complete patch included in the icedtea6-hg tree
> (http://icedtea.classpath.org/people/andrew/icedtea6-hg), including
> test case, is attached.

Tests are good of course and since the jaxp repo doesn't currently have 
any regression tests, putting the test into OpenJDK 6 in the jdk repo 
seems like the right home for it, as the patch currently does.

Note that at the moment the license header of the test files says 
"Copyright 2009 Red Hat" but later states "Please contact Sun 
Microsystems ... if you need additional information or have any 
questions."  If this goes upstream, the copyright holder should be 
changed to Sun.

Thanks,

-Joe