You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@subversion.apache.org by ep...@google.com on 2008/05/22 19:33:50 UTC

Re: svn commit: r31370 - trunk/subversion/bindings/javahl/native

arfrever@tigris.org writes:

> Author: arfrever
> Date: Thu May 22 12:27:52 2008
> New Revision: 31370
> 
> Log:
> * subversion/bindings/javahl/native/SVNClient.cpp
>   (SVNClient::getContext): Don't use KWallet in JavaHL.

Why not?

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: r31370 - trunk/subversion/bindings/javahl/native

Posted by Stefan Sperling <st...@elego.de>.
On Thu, May 22, 2008 at 07:33:20PM -0400, Mark Phippard wrote:
> On Thu, May 22, 2008 at 4:09 PM,  <ep...@google.com> wrote:
> > "Mark Phippard" <ma...@gmail.com> writes:
> >
> >> On Thu, May 22, 2008 at 3:48 PM,  <ep...@google.com> wrote:
> >> > "Mark Phippard" <ma...@gmail.com> writes:
> >> >
> >> >> If we wanted to support it in JavaHL for people in single-threaded
> >> >> environments I would need an easy way to disable it from Subclipse as
> >> >> the JVM crashes hard and fast as soon as that code gets executed.
> >> >
> >> > Just don't put that one in your auth providers list?
> >>
> >> That level of detail is not exposed to the JavaHL consumer.
> >
> > That is unfortunate.  Is the entire client context filled in for
> > you, or only this part?  How would you implement a configuration
> > option to disable certain auth providers?
> 
> Since I work via JavaHL, I do not necessarily know what the context
> is.  But I believe it includes things like the callbacks for
> authentication prompts and notification handlers.  Those are things we
> can provide.

In the command line client, providers are initialised in
libsvn_subr/cmdline.c, more precisely in svn_cmdline_setup_auth_baton.

It uses the svn_auth_get_*_provider functions to get the providers
it wants, puts them in a list, and passes that to svn_auth_open.

Doesn't JavaHL provide the svn_auth API? If it does not,
shouldn't JavaHL be extended so Java-based clients can handle
auth providers the same way the command line client does?

Stefan

Re: svn commit: r31370 - trunk/subversion/bindings/javahl/native

Posted by Mark Phippard <ma...@gmail.com>.
On Thu, May 22, 2008 at 4:09 PM,  <ep...@google.com> wrote:
> "Mark Phippard" <ma...@gmail.com> writes:
>
>> On Thu, May 22, 2008 at 3:48 PM,  <ep...@google.com> wrote:
>> > "Mark Phippard" <ma...@gmail.com> writes:
>> >
>> >> If we wanted to support it in JavaHL for people in single-threaded
>> >> environments I would need an easy way to disable it from Subclipse as
>> >> the JVM crashes hard and fast as soon as that code gets executed.
>> >
>> > Just don't put that one in your auth providers list?
>>
>> That level of detail is not exposed to the JavaHL consumer.
>
> That is unfortunate.  Is the entire client context filled in for
> you, or only this part?  How would you implement a configuration
> option to disable certain auth providers?

Since I work via JavaHL, I do not necessarily know what the context
is.  But I believe it includes things like the callbacks for
authentication prompts and notification handlers.  Those are things we
can provide.

-- 
Thanks

Mark Phippard
http://markphip.blogspot.com/

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: r31370 - trunk/subversion/bindings/javahl/native

Posted by ep...@google.com.
"Mark Phippard" <ma...@gmail.com> writes:

> On Thu, May 22, 2008 at 3:48 PM,  <ep...@google.com> wrote:
> > "Mark Phippard" <ma...@gmail.com> writes:
> >
> >> If we wanted to support it in JavaHL for people in single-threaded
> >> environments I would need an easy way to disable it from Subclipse as
> >> the JVM crashes hard and fast as soon as that code gets executed.
> >
> > Just don't put that one in your auth providers list?
> 
> That level of detail is not exposed to the JavaHL consumer.

That is unfortunate.  Is the entire client context filled in for
you, or only this part?  How would you implement a configuration
option to disable certain auth providers?

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: r31370 - trunk/subversion/bindings/javahl/native

Posted by Mark Phippard <ma...@gmail.com>.
On Thu, May 22, 2008 at 3:48 PM,  <ep...@google.com> wrote:
> "Mark Phippard" <ma...@gmail.com> writes:
>
>> If we wanted to support it in JavaHL for people in single-threaded
>> environments I would need an easy way to disable it from Subclipse as
>> the JVM crashes hard and fast as soon as that code gets executed.
>
> Just don't put that one in your auth providers list?

That level of detail is not exposed to the JavaHL consumer.

-- 
Thanks

Mark Phippard
http://markphip.blogspot.com/

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: r31370 - trunk/subversion/bindings/javahl/native

Posted by ep...@google.com.
"Mark Phippard" <ma...@gmail.com> writes:

> If we wanted to support it in JavaHL for people in single-threaded
> environments I would need an easy way to disable it from Subclipse as
> the JVM crashes hard and fast as soon as that code gets executed.

Just don't put that one in your auth providers list?

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org

Re: svn commit: r31370 - trunk/subversion/bindings/javahl/native

Posted by Mark Phippard <ma...@gmail.com>.
On Thu, May 22, 2008 at 3:33 PM,  <ep...@google.com> wrote:
> arfrever@tigris.org writes:
>
>> Author: arfrever
>> Date: Thu May 22 12:27:52 2008
>> New Revision: 31370
>>
>> Log:
>> * subversion/bindings/javahl/native/SVNClient.cpp
>>   (SVNClient::getContext): Don't use KWallet in JavaHL.
>
> Why not?

At least in terms of Subclipse's usage of JavaHL it crashes the JVM.
Eclipse uses a lot of threads.  SVNClient is initially initialized on
one thread but can then be used on any thread.  It is even very common
for SVN operations to run in the background and have to interact with
the UI thread.  It seems that the KDE code does not like this.  The
gnome-keyring works great as a point of reference.

I doubt it is related, but Eclipse uses gtk 2.0 so perhaps there is
some KDE initialization that does not happen when the application
starts?

If we wanted to support it in JavaHL for people in single-threaded
environments I would need an easy way to disable it from Subclipse as
the JVM crashes hard and fast as soon as that code gets executed.

-- 
Thanks

Mark Phippard
http://markphip.blogspot.com/

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org