Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
HDFS, mail # dev - Hadoop-2 generics compilation error


Copy link to this message
-
Re: Hadoop-2 generics compilation error
Andy Isaacson 2012-06-30, 00:12
Thanks everybody!  Good to get this resolved.

-andy

On Fri, Jun 29, 2012 at 3:53 PM, Alejandro Abdelnur <[EMAIL PROTECTED]>wrote:

> yep, just tested with sunjdk, and it works. We still have the
> unboxing, but is done when the variable is used as parameter in method
> that is expecting a primitive type.
>
>
>
> On Fri, Jun 29, 2012 at 3:49 PM, lars hofhansl <[EMAIL PROTECTED]>
> wrote:
> > SunJDK will correctly infer the reference types, correct?
> > If so, changing this to reference types should make both JDKs happy.
> >
> > -- Lars
> >
> > ________________________________
> > From: Alejandro Abdelnur <[EMAIL PROTECTED]>
> > To: [EMAIL PROTECTED]
> > Sent: Friday, June 29, 2012 3:26 PM
> > Subject: Re: Hadoop-2 generics compilation error
> >
> > Andy,
> >
> > It looks like OpenJDK javac does not handle type unboxing with a
> > generic method where the parameter that resolves the generic infers
> > the type to do the un-boxing on. While SunJDK does handle it.
> >
> > Now, to the question which one is correct? I'm not sure but SunJDK
> > does a valid inference.
> >
> > thx
> >
> > On Fri, Jun 29, 2012 at 2:43 PM, Andy Isaacson <[EMAIL PROTECTED]> wrote:
> >> I've filed https://issues.apache.org/jira/browse/HDFS-3580 and
> attached the
> >> patch.
> >>
> >> I'd really appreciate it if someone could explain what the problem is,
> why
> >> the patch fixes it, and whether this is a bug in OpenJDK (that it does
> not
> >> accept a valid program) or a bug in Sun JDK (that it does accept an
> >> incorrect program).
> >>
> >> -andy
> >>
> >> On Fri, Jun 29, 2012 at 12:01 PM, Andy Isaacson <[EMAIL PROTECTED]>
> wrote:
> >>
> >>> I ran into this as well, and am prepping a patch. Lars, if you could
> file
> >>> the Jira explaining why s/boolean/Boolean/ is the right fix, I'll test
> and
> >>> submit the patch.
> >>>
> >>> -andy
> >>>
> >>>
> >>> On Fri, Jun 29, 2012 at 11:08 AM, Alejandro Abdelnur <
> [EMAIL PROTECTED]>wrote:
> >>>
> >>>> This change is from HDFS-3481, the wrong JIRA number is because I made
> >>>> a mistake in the commit message.
> >>>>
> >>>> I can put a patch to fix this, but don't have OpenJDK to test it. It
> >>>> would be better if somebody with OpenJDK takes care of it and I'll
> >>>> review it.
> >>>>
> >>>> Thx
> >>>>
> >>>> On Fri, Jun 29, 2012 at 11:01 AM, Arun C Murthy <[EMAIL PROTECTED]>
> >>>> wrote:
> >>>> > HDFS-3113 is committed? Or, are you running with the patch yourself?
> >>>> >
> >>>> > It's still PA and I see Daryn reviewing it yet...
> >>>> >
> >>>> > Arun
> >>>> >
> >>>> > On Jun 28, 2012, at 10:27 PM, lars hofhansl wrote:
> >>>> >
> >>>> >> Since HDFS-3113 was integrated into Hadoop-2 I get the compilation
> >>>> errors of the following type:
> >>>> >>
> >>>> >> [ERROR]
> >>>>
> /home/lars/dev/hadoop-2/hadoop-hdfs-project/hadoop-hdfs-httpfs/src/main/java/org/apache/hadoop/fs/http/server/HttpFSServer.java:[407,36]
> >>>> incompatible types; no instance(s) of type variable(s) V exist so
> that V
> >>>> conforms to boolean
> >>>> >>
> >>>> >>
> >>>> >> Indeed at line 407 I see:
> >>>> >>
> >>>> >>         boolean hasData = params.get(DataParam.NAME,
> DataParam.class);
> >>>> >>
> >>>> >>
> >>>> >> When I change that to
> >>>> >>         Boolean hasData = params.get(DataParam.NAME,
> DataParam.class);
> >>>> >>
> >>>> >>
> >>>> >> instead (along with long to Long, short to Short, etc, later in
> that
> >>>> file), everything compiles fine.
> >>>> >>
> >>>> >> $ javac -version
> >>>> >> javac 1.6.0_24
> >>>> >>
> >>>> >>
> >>>> >> $ java -version
> >>>> >> java version "1.6.0_24"
> >>>> >> OpenJDK Runtime Environment (IcedTea6 1.11.3)
> >>>> (fedora-67.1.11.3.fc16-x86_64)
> >>>> >> OpenJDK 64-Bit Server VM (build 20.0-b12, mixed mode)
> >>>> >>
> >>>> >> Apparently OpenJDK gets mixed up on the type inference here. Can we
> >>>> change this in the way I suggest so that Hadoop-2 (and presumably
> trunk)
> >>>> can be
> >>>> >> compiled with OpenJDK?
> >>>> >> I'm happy to create a jira and (trivial) patch.