You are viewing a plain text version of this content. The canonical link for it is here.
Posted to java-dev@axis.apache.org by Steve Loughran <st...@iseran.com> on 2002/09/11 00:32:08 UTC
memory leaks
I've moved ant's storage of task references from strong refs to weak refs on
java1.2+, though we will have to see what breaks in the next Gump to work
out the full list of adverse consequences. Ideally, nothing, though the
conditions that forced the caching in the first place are pretty rare in
public projects.
-pull down the CVS_HEAD of jakarta-ant, "build" it (build install-lite is
fastest and puts it into ANT_HOME).,
-do an ant -version to make sure it is picked up
-do an ant all in axis.home.
-let me know of problems, either in Axis or anything else.
If the patch works ok it can be backfilled into ant1.5.1;
-steve
RE: memory leaks
Posted by Jon Ewing <jo...@mediasurface.com>.
<snip>
> but on a positive note, I didnt break the whole of the Gump
> nightly build with a barely tested patch to ant, which was a
> distinct possibility. I have stopped ant from building on
> java1.1, though, which I do need to correct.
>
> If it isnt Ant itself that is leaking, then the finger of
> suspicion has to go back to <javac>. Which may explain why
> I've never encountered the problem; i have
> build.compiler=jikes in my build.properties just for the
> speed difference; cleanup is a side effect. Everyone
> advocating continuous integration processes, including myself
> on p 396 of Java Dev with Ant, quite clearly say 'use jikes
> or fork javac'.
>
> I'm doing a build w/ javac instead of jikes to see if that
> can replicate the problem, but since I'm running java1.4.0, I
> have a different (not better, just different) version of
> javac. And I see that xml-axis on java.14 gump
> http://nagoya.apache.org/> ~rubys/gump/java14/xml-axis.html
>
> is failing with compile errors, not memory...
Well, downloaded the Axis CVS source first thing this morning, changed
nothing else, and it all compiled beautifully using javac.
jon.
Bug fix for org.apache.axis.message.SOAPBody
Posted by Allegar Robert <al...@bah.com>.
The method void addBodyElement(SOAPBodyElement element) in
org.apache.axis.message.SOAPBody is missing the following line of code
around line 170 in the 9.11.2002 build:
bodyElements.addElement(element);
It doesn't actually add the passed-in element to the vector of body elements
without it. I don't have cvs or diff installed so apologies if the format
for this fix is incorrect.
Regards,
Rob
-----Original Message-----
From: Steve Loughran [mailto:steve_l@iseran.com]
Sent: Wednesday, September 11, 2002 1:33 PM
To: axis-dev@xml.apache.org; dims@yahoo.com
Subject: Re: memory leaks
----- Original Message -----
From: "Davanum Srinivas" <di...@yahoo.com>
To: <ax...@xml.apache.org>
Sent: Wednesday, September 11, 2002 10:58 AM
Subject: Re: memory leaks
> http://developer.java.sun.com/developer/bugParade/bugs/4482232.html
>
oh, I love that comment
" This program appears to be using an unsupported internal interface.
Why don't you build using javac like everybody else?", it's right up there
with "you dont need unsigned datatypes".
The irony is that axis uses javac in process for .jws pages, as do most jsp
runtimes. Hey, even jasper, the reference implementation of jsp that Sun
helped write uses javac inproc. Though tomcat 5.0 will hand off the builds
to ant. The other irony is that CodeDOM in .NET not only makes it possible
to compile code inproc, it makes it easy to create it.
Is there any way to run the tests using jikes, even if we hang on to javac
for axis.jar itself, on account of generated code quality?
Re: memory leaks
Posted by Steve Loughran <st...@iseran.com>.
----- Original Message -----
From: "Davanum Srinivas" <di...@yahoo.com>
To: <ax...@xml.apache.org>
Sent: Wednesday, September 11, 2002 10:58 AM
Subject: Re: memory leaks
> http://developer.java.sun.com/developer/bugParade/bugs/4482232.html
>
oh, I love that comment
" This program appears to be using an unsupported internal interface.
Why don't you build using javac like everybody else?", it's right up there
with "you dont need unsigned datatypes".
The irony is that axis uses javac in process for .jws pages, as do most jsp
runtimes. Hey, even jasper, the reference implementation of jsp that Sun
helped write uses javac inproc. Though tomcat 5.0 will hand off the builds
to ant. The other irony is that CodeDOM in .NET not only makes it possible
to compile code inproc, it makes it easy to create it.
Is there any way to run the tests using jikes, even if we hang on to javac
for axis.jar itself, on account of generated code quality?
Re: memory leaks
Posted by Davanum Srinivas <di...@yahoo.com>.
http://developer.java.sun.com/developer/bugParade/bugs/4482232.html
--- Steve Loughran <st...@iseran.com> wrote:
>
> ----- Original Message -----
> From: "jon ewing" <jo...@mediasurface.com>
> To: <ax...@xml.apache.org>
> Sent: Wednesday, September 11, 2002 10:18 AM
> Subject: RE: memory leaks
>
>
> >
> > > -----Original Message-----
> > > From: Steve Loughran [mailto:steve_l@iseran.com]
> > > Sent: 11 September 2002 17:36
> > > To: axis-dev@xml.apache.org
> > > Subject: Re: memory leaks
> > >
> > >
> > >
> > > ----- Original Message -----
> > > From: "Jon Ewing" <jo...@mediasurface.com>
> > > To: <ax...@xml.apache.org>
> > > Sent: Wednesday, September 11, 2002 2:45 AM
> > > Subject: RE: memory leaks
> > >
> > >
> > > >
> > > > Hi,
> > > >
> > > > Downloaded and compiled Ant from CVS source this morning:
> > > >
> > > > % ant -v
> > > > Apache Ant version 1.6alpha compiled on September 11 2002
> > > >
> > > > But
> > > >
> > > > % ant functional-tests
> > > >
> > > > in yesterday's CVS tree of Axis still bombed in the same
> > > place with a
> > > > java.lang.OutOfMemoryError (Java 1.3.1_04, Solaris 8).
> > > >
> > > > Jon.
> > >
> > >
> > > bugger.
> >
> > My thoughts exactly ;)
>
> but on a positive note, I didnt break the whole of the Gump nightly build
> with a barely tested patch to ant, which was a distinct possibility. I have
> stopped ant from building on java1.1, though, which I do need to correct.
>
> If it isnt Ant itself that is leaking, then the finger of suspicion has to
> go back to <javac>. Which may explain why I've never encountered the
> problem; i have build.compiler=jikes in my build.properties just for the
> speed difference; cleanup is a side effect. Everyone advocating continuous
> integration processes, including myself on p 396 of Java Dev with Ant, quite
> clearly say 'use jikes or fork javac'.
>
> I'm doing a build w/ javac instead of jikes to see if that can replicate the
> problem, but since I'm running java1.4.0, I have a different (not better,
> just different) version of javac. And I see that xml-axis on java.14 gump
> http://nagoya.apache.org/~rubys/gump/java14/xml-axis.html
>
> is failing with compile errors, not memory...
>
>
=====
Davanum Srinivas - http://xml.apache.org/~dims/
__________________________________________________
Yahoo! - We Remember
9-11: A tribute to the more than 3,000 lives lost
http://dir.remember.yahoo.com/tribute
Re: memory leaks
Posted by Steve Loughran <st...@iseran.com>.
----- Original Message -----
From: "jon ewing" <jo...@mediasurface.com>
To: <ax...@xml.apache.org>
Sent: Wednesday, September 11, 2002 10:18 AM
Subject: RE: memory leaks
>
> > -----Original Message-----
> > From: Steve Loughran [mailto:steve_l@iseran.com]
> > Sent: 11 September 2002 17:36
> > To: axis-dev@xml.apache.org
> > Subject: Re: memory leaks
> >
> >
> >
> > ----- Original Message -----
> > From: "Jon Ewing" <jo...@mediasurface.com>
> > To: <ax...@xml.apache.org>
> > Sent: Wednesday, September 11, 2002 2:45 AM
> > Subject: RE: memory leaks
> >
> >
> > >
> > > Hi,
> > >
> > > Downloaded and compiled Ant from CVS source this morning:
> > >
> > > % ant -v
> > > Apache Ant version 1.6alpha compiled on September 11 2002
> > >
> > > But
> > >
> > > % ant functional-tests
> > >
> > > in yesterday's CVS tree of Axis still bombed in the same
> > place with a
> > > java.lang.OutOfMemoryError (Java 1.3.1_04, Solaris 8).
> > >
> > > Jon.
> >
> >
> > bugger.
>
> My thoughts exactly ;)
but on a positive note, I didnt break the whole of the Gump nightly build
with a barely tested patch to ant, which was a distinct possibility. I have
stopped ant from building on java1.1, though, which I do need to correct.
If it isnt Ant itself that is leaking, then the finger of suspicion has to
go back to <javac>. Which may explain why I've never encountered the
problem; i have build.compiler=jikes in my build.properties just for the
speed difference; cleanup is a side effect. Everyone advocating continuous
integration processes, including myself on p 396 of Java Dev with Ant, quite
clearly say 'use jikes or fork javac'.
I'm doing a build w/ javac instead of jikes to see if that can replicate the
problem, but since I'm running java1.4.0, I have a different (not better,
just different) version of javac. And I see that xml-axis on java.14 gump
http://nagoya.apache.org/~rubys/gump/java14/xml-axis.html
is failing with compile errors, not memory...
RE: memory leaks
Posted by jon ewing <jo...@mediasurface.com>.
> -----Original Message-----
> From: Steve Loughran [mailto:steve_l@iseran.com]
> Sent: 11 September 2002 17:36
> To: axis-dev@xml.apache.org
> Subject: Re: memory leaks
>
>
>
> ----- Original Message -----
> From: "Jon Ewing" <jo...@mediasurface.com>
> To: <ax...@xml.apache.org>
> Sent: Wednesday, September 11, 2002 2:45 AM
> Subject: RE: memory leaks
>
>
> >
> > Hi,
> >
> > Downloaded and compiled Ant from CVS source this morning:
> >
> > % ant -v
> > Apache Ant version 1.6alpha compiled on September 11 2002
> >
> > But
> >
> > % ant functional-tests
> >
> > in yesterday's CVS tree of Axis still bombed in the same
> place with a
> > java.lang.OutOfMemoryError (Java 1.3.1_04, Solaris 8).
> >
> > Jon.
>
>
> bugger.
My thoughts exactly ;)
Re: memory leaks
Posted by Steve Loughran <st...@iseran.com>.
----- Original Message -----
From: "Jon Ewing" <jo...@mediasurface.com>
To: <ax...@xml.apache.org>
Sent: Wednesday, September 11, 2002 2:45 AM
Subject: RE: memory leaks
>
> Hi,
>
> Downloaded and compiled Ant from CVS source this morning:
>
> % ant -v
> Apache Ant version 1.6alpha compiled on September 11 2002
>
> But
>
> % ant functional-tests
>
> in yesterday's CVS tree of Axis still bombed in the same place
> with a java.lang.OutOfMemoryError (Java 1.3.1_04, Solaris 8).
>
> Jon.
bugger.
Re: memory leaks
Posted by Steve Loughran <st...@iseran.com>.
----- Original Message -----
From: "Davanum Srinivas" <di...@yahoo.com>
To: <ax...@xml.apache.org>
Sent: Wednesday, September 11, 2002 9:42 AM
Subject: Re: memory leaks
> Steve,
>
> Gump Failed...It was not logged correctly - See
> http://jakarta.apache.org/builds/gump/latest/xml-axis.html for details.
>
> Thanks,
> dims
yeah, It did seem to trail off towards the end. But the summary looked
happy.
>
> PS: I added a fork option to foreach today. Can you please tweak it? i
don't have too much
> experience with Ant.
will do
Re: memory leaks
Posted by Davanum Srinivas <di...@yahoo.com>.
Steve,
Gump Failed...It was not logged correctly - See
http://jakarta.apache.org/builds/gump/latest/xml-axis.html for details.
Thanks,
dims
PS: I added a fork option to foreach today. Can you please tweak it? i don't have too much
experience with Ant.
--- Steve Loughran <st...@iseran.com> wrote:
>
> ----- Original Message -----
> From: "Jon Ewing" <jo...@mediasurface.com>
> To: <ax...@xml.apache.org>
> Sent: Wednesday, September 11, 2002 2:45 AM
> Subject: RE: memory leaks
>
>
> >
> > Hi,
> >
> > Downloaded and compiled Ant from CVS source this morning:
> >
> > % ant -v
> > Apache Ant version 1.6alpha compiled on September 11 2002
> >
> > But
> >
> > % ant functional-tests
> >
> > in yesterday's CVS tree of Axis still bombed in the same place
> > with a java.lang.OutOfMemoryError (Java 1.3.1_04, Solaris 8).
> >
>
> hmm. something else may be going wrong. It looks like Axis succeeded last
> night, but enough changes were made to <foreach> and other parts of the
> build that the leak has been pushed back a bit, rather than eliminated.
>
> I wonder if I could create an ant unit test to replicate the problem, so
> that it could be tracked down better.
>
>
=====
Davanum Srinivas - http://xml.apache.org/~dims/
__________________________________________________
Yahoo! - We Remember
9-11: A tribute to the more than 3,000 lives lost
http://dir.remember.yahoo.com/tribute
Re: memory leaks
Posted by Steve Loughran <st...@iseran.com>.
----- Original Message -----
From: "Jon Ewing" <jo...@mediasurface.com>
To: <ax...@xml.apache.org>
Sent: Wednesday, September 11, 2002 2:45 AM
Subject: RE: memory leaks
>
> Hi,
>
> Downloaded and compiled Ant from CVS source this morning:
>
> % ant -v
> Apache Ant version 1.6alpha compiled on September 11 2002
>
> But
>
> % ant functional-tests
>
> in yesterday's CVS tree of Axis still bombed in the same place
> with a java.lang.OutOfMemoryError (Java 1.3.1_04, Solaris 8).
>
hmm. something else may be going wrong. It looks like Axis succeeded last
night, but enough changes were made to <foreach> and other parts of the
build that the leak has been pushed back a bit, rather than eliminated.
I wonder if I could create an ant unit test to replicate the problem, so
that it could be tracked down better.
RE: memory leaks
Posted by Jon Ewing <jo...@mediasurface.com>.
Hi,
Downloaded and compiled Ant from CVS source this morning:
% ant -v
Apache Ant version 1.6alpha compiled on September 11 2002
But
% ant functional-tests
in yesterday's CVS tree of Axis still bombed in the same place
with a java.lang.OutOfMemoryError (Java 1.3.1_04, Solaris 8).
Jon.
> -----Original Message-----
> From: Steve Loughran [mailto:steve_l@iseran.com]
> Sent: 10 September 2002 23:32
> To: axis-dev
> Subject: memory leaks
>
>
> I've moved ant's storage of task references from strong refs
> to weak refs on java1.2+, though we will have to see what
> breaks in the next Gump to work out the full list of adverse
> consequences. Ideally, nothing, though the conditions that
> forced the caching in the first place are pretty rare in
> public projects.
>
> -pull down the CVS_HEAD of jakarta-ant, "build" it (build
> install-lite is fastest and puts it into ANT_HOME)., -do an
> ant -version to make sure it is picked up -do an ant all in
> axis.home. -let me know of problems, either in Axis or anything else.
>
>
> If the patch works ok it can be backfilled into ant1.5.1;
>
> -steve
>