geode-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Jacob Barrett <jbarr...@pivotal.io>
Subject Re: [Discuss] Compatibility with Apache Xerces
Date Mon, 21 Aug 2017 16:09:21 GMT
Looking into the legacy repo I can see that the fix for the StringBuffer
issue was not brought into Geode. The fix was to specifically request the
Oracle/Sun implementation. Since that implementation might not be available
in the IBM JDK its probably not a great fix.


On Fri, Aug 18, 2017 at 9:28 PM Darren Foong <darrenfoong@gmail.com> wrote:

> > You're right. This did come up with the IBM JDK and we fixed it. Not
> sure why then it's coming up again.
>
> Could it be this? https://issues.apache.org/jira/browse/GEODE-135
> Seems like a different issue with the IBM JDK though.
>
> I've just verified that the whitespace problem is still an issue with
> the IBM JDK (because it uses Apache Xerces), and this pull request
> fixes it.
>
> Not sure why the reporter for GEODE-135 didn't face a similar problem back
> then.
>
> - Darren
>
> On Sat, Aug 19, 2017 at 6:04 AM, Jacob Barrett <jbarrett@pivotal.io>
> wrote:
> > You're right. This did come up with the IBM JDK and we fixed it. Not
> sure why then it's coming up again.
> >
> > Sent from my iPhone
> >
> > On Aug 18, 2017, at 2:01 PM, Dan Smith <dsmith@pivotal.io> wrote:
> >
> >>> Since the oracle parser is always going to be there I don't see any
> harm
> >> in doing that.
> >>
> >> That's not true if we're running on non-oracle JDKs, right? I remember a
> >> while back someone was trying to run geode on IBMs JDK and having
> issues -
> >> maybe even this same whitespace problem?
> >>
> >> I think it this fixes issues with other parsers it looks good to me, I
> >> don't have a problem with adding xerces as a test dependency.
> >>
> >> -Dan
> >>
> >>> On Fri, Aug 18, 2017 at 10:48 AM, Jacob Barrett <jbarrett@pivotal.io>
> wrote:
> >>>
> >>> I could have sworn at one point the the cache xml parser explicitly
> >>> requested the oracle parser. Since the oracle parser is always going
> to be
> >>> there I don't see any harm in doing that.
> >>>
> >>> A better fix might be to just normalize the white space when parsing.
> >>>
> >>> I also recall xerces having a flag for controlling the white space
> >>> treatment.
> >>>
> >>> -Jake
> >>>
> >>>
> >>> Sent from my iPhone
> >>>
> >>>> On Aug 18, 2017, at 10:25 AM, Anilkumar Gingade <agingade@pivotal.io>
> >>> wrote:
> >>>>
> >>>> Why worry is claiming to support multiple version; and trying to
> >>>> manage/maintain it...
> >>>>
> >>>> -Anil.
> >>>>
> >>>>
> >>>> On Thu, Aug 17, 2017 at 11:35 PM, Darren Foong <darrenfoong@gmail.com
> >
> >>>> wrote:
> >>>>
> >>>>> Hi all,
> >>>>>
> >>>>> I'm using Geode in an application that uses the Apache implementation
> >>>>> of Xerces. The Oracle JDK comes with its own implementation of
> Xerces.
> >>>>>
> >>>>> I encountered an issue
> >>>>> (https://issues.apache.org/jira/browse/GEODE-3306) whereby cache.xml
> >>>>> parsing fails with Apache Xerces; details are in JIRA.
> >>>>>
> >>>>> Currently there are two workarounds:
> >>>>>
> >>>>> 1. Remove the whitespace between elements in cache.xml
> >>>>> 2. Load the JDK Xerces when parsing cache.xml
> >>>>>
> >>>>> I've submitted a pull request
> >>>>> (https://github.com/apache/geode/pull/668) to make `CacheXmlParser`
> >>>>> compatible with both versions of Xerces.
> >>>>>
> >>>>> This change would be useful for at least two groups of people:
> >>>>>
> >>>>> 1. Developers who are using the Apache implementation of Xerces
> >>>>> throughout their application, and only want one implementation of
> >>>>> Xerces
> >>>>> 2. Developers who are using a non-Oracle JDK
> >>>>>
> >>>>> Does anyone have any objections to having `xercesImpl` as a test
> >>>>> runtime dependency?
> >>>>>
> >>>>> I'd appreciate any feedback. Thank you!
> >>>>>
> >>>>> Best regards,
> >>>>> - Darren Foong
> >>>>>
> >>>
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message