apologies, 3.1, mingw wishlist

apologies, 3.1, mingw wishlist

[Adam Megacz]

First off, sorry I've been neglecting gcj recently. I needed some time
off from coding.

Second: is 3.1 still scheduled for release on 15-Apr, or has it been
pushed back?

Finally, I'm going to have a block of time Thursday through Sunday for
more Win32 work. Right now my goals are definately:

    - Fix the badly-broken notifyAll()
    - Make Win32Process work
    - Hash Synchronization (this already works, I just have to merge patches)
    - Closer review of java.io.* for proper error handling

And, if I have time:

    - Stack traces
    - Get the testsuite to run (somehow -- not sure yet how I'll do it)

Is there anything else that I should consider? After this weekend, it
might be a while before I have time to visit this again, so if there's
something you want to see, please let me know now... =)

  - a

-- 
The web is dead; long live the Internet.
http://www.xwt.org/

Re: apologies, 3.1, mingw wishlist

[Tom Tromey]

>>>>> "Adam" == Adam Megacz <gcj@lists.megacz.com> writes:

Adam> Second: is 3.1 still scheduled for release on 15-Apr, or has it
Adam> been pushed back?

I assume it has been pushed back but I have not heard either official
or unofficial word.

Adam> - Fix the badly-broken notifyAll()
Adam> - Make Win32Process work
Adam> - Hash Synchronization (this already works, I just have to merge patches)
Adam> - Closer review of java.io.* for proper error handling

These all sound great.

Depending on timing, etc, they might not make it into 3.1.  That isn't
fatal.  We could always get the changes into 3.1.1.

Adam> - Get the testsuite to run (somehow -- not sure yet how I'll do it)

AG suggests using Wine.  Anything that works is fine by me though.

Adam> Is there anything else that I should consider?

You know the Windows status better than anybody.  How about submitting
a PR for all the problems and missing things you know of once you've
finished the next round of coding?  That way we'll have a better idea
of the state and to-do items later on.

Tom

Re: apologies, 3.1, mingw wishlist

[Brian Jones]

Tom Tromey <tromey@redhat.com> writes:

> Adam> - Get the testsuite to run (somehow -- not sure yet how I'll do it)
> 
> AG suggests using Wine.  Anything that works is fine by me though.

If you have to, running Mauve under cygwin shouldn't be too difficult.

Brian
-- 
Brian Jones <cbj@gnu.org>

Re: apologies, 3.1, mingw wishlist

[Adam King]

On Tue, Apr 09, 2002 at 03:20:20PM -0700, Adam Megacz wrote:
> 
> First off, sorry I've been neglecting gcj recently. 

Don't be - you did a great job supporting the mingw32 target!
Getting our 200k+ lines of java code to work with it is progressing
much better than I expected.  So far, I've only come across three
places where we use an API method not supported by gcj.

If you're taking votes... 

>     - Make Win32Process work
Yes please!  We use it in a couple of places, and if you don't
add it, I'll have to in a couple of weeks.  Also, if you want to
add in Runtime.exec( String[], String[], File ) (added in 1.3 I think )
that would be great.

> And, if I have time:
> 
>     - Stack traces

This would be _very_ useful - while getting our code to compile, an 
exception would be thrown but I didn't know from where.  I'd end up
having to stick a bunch of printlns throughout the code to track
down its cause.

> Is there anything else that I should consider? 

Is there any way that GDB would/could be used for debugging?  

Anyways, thanks again for your work on the port ( and also, to
all the gcj coders! !

	Adam K

Re: apologies, 3.1, mingw wishlist [okay to add new ctor to ConcreteProcess?]

[Adam Megacz]

Adam King <aking@dreammechanics.com> writes:
> add it, I'll have to in a couple of weeks.  Also, if you want to
> add in Runtime.exec( String[], String[], File ) (added in 1.3 I think )
> that would be great.

Sure, if it's okay with everybody else, I'll create another
constructor on ConcreteProcess that takes a current-working-directory
argument.



> > And, if I have time:
> >     - Stack traces

> This would be _very_ useful - while getting our code to compile, an 
> exception would be thrown but I didn't know from where.  I'd end up
> having to stick a bunch of printlns throughout the code to track
> down its cause.

Yeah, looking back (and with the help of others on the list who showed
me how easy it can be), I spent more time putting in printf()'s for
XWT than it would have taken to just write the stack trace code and be
done with it.


> > Is there anything else that I should consider? 

> Is there any way that GDB would/could be used for debugging?  

Hrm, I already use cygwin's GDB. It's not the best -- if your code
CALL's a NULL address you lose the stack, and the GUI is a bit lacking
-- but it's better than nothing.

 - a


-- 
The web is dead; long live the Internet.
http://www.xwt.org/

Re: apologies, 3.1, mingw wishlist [okay to add new ctor to ConcreteProcess?]

[Tom Tromey]

>>>>> "Adam" == Adam Megacz <gcj@lists.megacz.com> writes:

Adam> Sure, if it's okay with everybody else, I'll create another
Adam> constructor on ConcreteProcess that takes a
Adam> current-working-directory argument.

Probably you want to make that the only constructor and change Runtime
to always use it.

>> Is there any way that GDB would/could be used for debugging?  

Adam> Hrm, I already use cygwin's GDB. It's not the best -- if your
Adam> code CALL's a NULL address you lose the stack, and the GUI is a
Adam> bit lacking -- but it's better than nothing.

The current gdb's gcj support isn't all that great.
Even worse, nobody is working on it :-(

Tom

Java debugger

[Adam King]

On Thu, Apr 11, 2002 at 03:28:37PM -0600, Tom Tromey wrote:
> 
> Adam M> Hrm, I already use cygwin's GDB. It's not the best -- if your
> Adam M> code CALL's a NULL address you lose the stack, and the GUI is a
> Adam M> bit lacking -- but it's better than nothing.
> 
> The current gdb's gcj support isn't all that great.
> Even worse, nobody is working on it :-(

hmm.. I currently use JSwat for most of my current java debugging
needs - so I was wondering, is anyone working on a JPDA 
implementation?  Would it even be possible since it's something
I wouldn't mind giving a try adding to gcj.

	Adam

Re: Java debugger

[Nic Ferrier]

Adam King <aking@dreammechanics.com> writes:

> On Thu, Apr 11, 2002 at 03:28:37PM -0600, Tom Tromey wrote: 
> >  
> > Adam M> Hrm, I already use cygwin's GDB. It's not the best -- if your 
> > Adam M> code CALL's a NULL address you lose the stack, and the GUI is a 
> > Adam M> bit lacking -- but it's better than nothing. 
> >  
> > The current gdb's gcj support isn't all that great. 
> > Even worse, nobody is working on it :-( 
>  
> hmm.. I currently use JSwat for most of my current java debugging 
> needs - so I was wondering, is anyone working on a JPDA  
> implementation?  Would it even be possible since it's something 
> I wouldn't mind giving a try adding to gcj. 

GDB is the debugger of choice for gcj hackers. It's a shame that GCJ
still can't debug byte codes because compiling lots of code to native
code can be a pain.


Nic

Re: Java debugger

[Bryce McKinlay]

Adam King wrote:

>On Thu, Apr 11, 2002 at 03:28:37PM -0600, Tom Tromey wrote:
>
>>Adam M> Hrm, I already use cygwin's GDB. It's not the best -- if your
>>Adam M> code CALL's a NULL address you lose the stack, and the GUI is a
>>Adam M> bit lacking -- but it's better than nothing.
>>
>>The current gdb's gcj support isn't all that great.
>>Even worse, nobody is working on it :-(
>>
>
>hmm.. I currently use JSwat for most of my current java debugging
>needs - so I was wondering, is anyone working on a JPDA 
>implementation?  Would it even be possible since it's something
>I wouldn't mind giving a try adding to gcj.
>

Its something I've been thinking about too and is definatly doable, I'd 
say, with a combination of ptrace() and a DWARF2 interpreter. One 
problem is that if we wanted JPDA to be part of libgcj itself we cannot 
use or link any BFD or GDB code due to the license. But, if we can't 
live without code from those, perhaps we could make the JPDA back-end a 
separate, GPLed module which plugs in to libgcj somehow.

Then again, if we were to write the interpreter ourselves we might also 
be able to get away without using the BFD stuff, but I can imagine it 
could require a lot more work to port it to various different 
(especially non-ELF) platforms.

regards

Bryce.

Re: apologies, 3.1, mingw wishlist [okay to add new ctor to ConcreteProcess?]

[Adam Megacz]

Tom Tromey <tromey@redhat.com> writes:
> Probably you want to make that the only constructor and change Runtime
> to always use it.

Even better.

  - a

-- 
The web is dead; long live the Internet.
http://www.xwt.org/