[RFC][icedtea6-1.9, icedtea6-1.10] Backport for PR732: Use xsltproc for bootstrap xslt in place of Xerces/Xalan

[Andrew Hughes]

----- Original Message -----
> On 21/12/11 10:25 AM, Deepak Bhole wrote:
> > * Dr Andrew John Hughes<ahughes at redhat.com>  [2011-12-21 10:23]:
> >> On 14:57 Tue 20 Dec     , Danesh Dadachanji wrote:
> >>> Adding CC. =)
> >>>
> >>> On 15/11/11 04:54 PM, Danesh Dadachanji wrote:
> >>>> Ping?
> >>>>
> >>>> On 26/10/11 04:13 PM, Danesh Dadachanji wrote:
> >>>>> Hi,
> >>>>>
> >>>>> Attached are patches for 6-1.9 and 6-1.10 to replace the use of
> >>>>> xalan/xerces with xsltproc.
> >>>>>
> >>>>> It is pretty similar to PR732's fix[1] with some minor changes
> >>>>> in
> >>>>> syntax/line numbers for diffs:
> >>>>> - The changeset from HEAD that added
> >>>>> patches/ecj/endorsed-dir-for-jvmti.patch was never backported
> >>>>> so I did
> >>>>> not need to deal with it.
> >>>>> - I had to adjust patches/ecj/xsltproc.patch to get it to
> >>>>> apply,
> >>>>> jvmti.make is a bit offset when compared to HEAD's build.
> >>>>> - For 1.9, the configure.ac macros in question do not start
> >>>>> with "IT_"
> >>>>> (e.g. FIND_TOOL vs IT_FIND_TOOL).
> >>>>>
> >>>>> 1.10 ChangeLog:
> >>>>> +2011-10-26 Danesh Dadachanji<ddadacha at redhat.com>
> >>>>> +
> >>>>> + * Makefile.am:
> >>>>> + (ENDORSED_DIR): Removed.
> >>>>> + (ICEDTEA_ECJ_PATCHES): Add xsltproc.patch.
> >>>>> + (ICEDTEA_ENV_ECJ): Set XSLT. Don't
> >>>>> + set ICEDTEA_ENDORSED_DIR or ENDORSED.
> >>>>> + (bootstrap-directory-stage1): Don't link
> >>>>> + xalan2, xerces2 or xalan2 serialiser.
> >>>>> + * acinclude.m4:
> >>>>> + (IT_FIND_XALAN2_JAR): Removed.
> >>>>> + (IT_FIND_XALAN2_SERIALIZER_JAR): Removed.
> >>>>> + (IT_FIND_XERCES2_JAR): Removed.
> >>>>> + * configure.ac:
> >>>>> + Replace calls to FIND_XALAN2_JAR,
> >>>>> + IT_FIND_XALAN2_SERIALIZER_JAR and
> >>>>> + IT_FIND_XERCES2_JAR with call to
> >>>>> + IT_FIND_TOOL for xsltproc.
> >>>>> + * patches/ecj/xsltproc.patch:
> >>>>> + Patch jvmti.make in HotSpot to use
> >>>>> + xsltproc instead of Java's XSLT
> >>>>> + when bootstrapping.
> >>>>> + * NEWS: Updated.
> >>>>>
> >>>>>
> >>>>> 1.9 ChangeLog:
> >>>>> +2011-10-26 Danesh Dadachanji<ddadacha at redhat.com>
> >>>>> +
> >>>>> + * Makefile.am:
> >>>>> + (ENDORSED_DIR): Removed.
> >>>>> + (ICEDTEA_ECJ_PATCHES): Add xsltproc.patch.
> >>>>> + (ICEDTEA_ENV_ECJ): Set XSLT. Don't
> >>>>> + set ICEDTEA_ENDORSED_DIR or ENDORSED.
> >>>>> + (bootstrap-directory-stage1): Don't link
> >>>>> + xalan2, xerces2 or xalan2 serialiser.
> >>>>> + * acinclude.m4:
> >>>>> + (FIND_XALAN2_JAR): Removed.
> >>>>> + (FIND_XALAN2_SERIALIZER_JAR): Removed.
> >>>>> + (FIND_XERCES2_JAR): Removed.
> >>>>> + * configure.ac:
> >>>>> + Replace calls to FIND_XALAN2_JAR,
> >>>>> + FIND_XALAN2_SERIALIZER_JAR and
> >>>>> + FIND_XERCES2_JAR with call to
> >>>>> + FIND_TOOL for xsltproc.
> >>>>> + * patches/ecj/xsltproc.patch:
> >>>>> + Patch jvmti.make in HotSpot to use
> >>>>> + xsltproc instead of Java's XSLT
> >>>>> + when bootstrapping.
> >>>>> + * NEWS: Updated.
> >>>>>
> >>>>> Regards,
> >>>>> Danesh
> >>>>> [1] http://icedtea.classpath.org/hg/icedtea6/rev/32fa8c401cee
> >>
> >> What's the status of this?
> 
> It's just here in review, I haven't pushed anything yet. Testing
> builds
> on F15 have worked fine (where before they crashed with lack of
> xalan/xerces access).
> 
> >>
> >> I don't think we should make such extensive build changes on a
> >> release branch.
> >
> > Agreed. What is the motivation for putting this in a re;ease
> > branch?
> 
> The motivation was that these branches fail to build on F15 onward.
> If
> you build a fresh clone of either release branch, you'd run into the
> xalan/xerces dependency resolution bug. Some members of the community
> do
> not know the workaround (patch that adds a symlink of
> xml-commons-apis.jar in endorsed dir) so Andrew Hughes asked me to
> backport his patch.
> 
> As mentioned in the original email, the changes from the patch to
> HEAD
> are very minimal, just refactoring/removing redundant line deletions.
> It's been in HEAD for quite some time now too so I believe it has
> been
> soaked in fine and should be good to backport. However, I know the
> general terms are that backports should be made if users are affected
> and this is more in favour of devs. I have no problems scrapping it
> if
> you think there is no need.
> 
> Regards,
> Danesh
> 

I usually don't like to make build changes to release branches, but I think
we can make an exception here.  Keeping the status quo causing user headaches
(circular dependencies due to pulling in xerces/xalan, plus an additional issue
on F15+ due to changes in recent versions) while the amount of change in the patch
and our experience on HEAD for both 6 & 7 (where it's been present for even longer)
suggests this should be ok to apply.

Unless anyone else has any major objections to this, I'd like to see this applied
after the security update, so it forms part of 1.9.14 and 1.10.7.

Please make sure the change is mentioned in NEWS (I didn't notice whether it was or not in the patch).
-- 
Andrew :)

Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

PGP Key: 248BDC07 (https://keys.indymedia.org/)
Fingerprint = EC5A 1F5E C0AD 1D15 8F1F  8F91 3B96 A578 248B DC07