gij and -cp/-classpath

gij and -cp/-classpath

[Andrew Pinski]

Would you accept a patch for this?
Also would some give an hit of how to do it?

Mainly the work needs to be done so `gij' would work like `java'.

I would need help on the documentation.

Thanks,
Andrew Pinski

Re: gij and -cp/-classpath

[Tom Tromey]

>>>>> "Andrew" == Andrew Pinski <pinskia@physics.uc.edu> writes:

Andrew> Would you accept a patch for this?

Yes, definitely.

Andrew> Also would some give an hit of how to do it?

At one point Nic was revamping the argument processing, but I don't
know the status of this.

Meanwhile you can do it by editing libjava/gij.cc.  Any option will
require new code in the help() function plus some processing in main().

The class path is actually set in natSystem.cc.  Getting the
information from gij to here is a little tricky.  Maybe a new global
is best.

You also need to figure out how the new option interacts with the
CLASSPATH environment variable and with `-jar' mode.

Andrew> Mainly the work needs to be done so `gij' would work like
Andrew> `java'.

This is definitely a goal of ours.  Any work towards this is
appreciated.

Andrew> I would need help on the documentation.

The docs are in gcc/java/gcj.texi.  If you can write text describing
what the new option(s) do, I will add the texinfo markup for you.

Tom

Re: gij and -cp/-classpath

[Nic Ferrier]

Tom Tromey <tromey@redhat.com> writes:

> >>>>> "Andrew" == Andrew Pinski <pinskia@physics.uc.edu> writes: 
>  
> Andrew> Would you accept a patch for this? 
>  
> Yes, definitely. 
>  
> Andrew> Also would some give an hit of how to do it? 
>  
> At one point Nic was revamping the argument processing, but I don't 
> know the status of this. 

I did write an argument parser and an implementation of gij based on
the argument parser. I never submitted it because there wasn't enough
enthusiasm for it.

The GCJ folks mainly seem to be interested in a change to the
CNI/JNI invocation API. I refocussed my efforts on that but haven't
had time to come up with a definitive patch (I am still hoping to do
that however).


Nic

Re: gij and -cp/-classpath

[Bryce McKinlay]

Nic Ferrier wrote:

>The GCJ folks mainly seem to be interested in a change to the
>CNI/JNI invocation API. I refocussed my efforts on that but haven't
>had time to come up with a definitive patch (I am still hoping to do
>that however).
>

Yes, the idea is to have one argument parser which can be shared by:

- gij command line
- vm_args for CNI & JNI invocation API
- an environment variable (LIBGCJ_ARGS or some such) which sets VM 
parameters for any gcj binary

regards

Bryce.

Re: gij and -cp/-classpath

[Nic Ferrier]

Bryce McKinlay <bryce@waitaki.otago.ac.nz> writes:

> Yes, the idea is to have one argument parser which can be shared by: 
>  
> - gij command line 
> - vm_args for CNI & JNI invocation API 
> - an environment variable (LIBGCJ_ARGS or some such) which sets VM  
> parameters for any gcj binary 

I don't understand that last part: don't *all* gcj applications have
to use the invocation API? I presumed that GCJ compiled (native) code
was wrapped with some invocation caller.


Nic

Re: gij and -cp/-classpath

[Bryce McKinlay]

Nic Ferrier wrote:

>Bryce McKinlay <bryce@waitaki.otago.ac.nz> writes:
>
>>Yes, the idea is to have one argument parser which can be shared by: 
>> 
>>- gij command line 
>>- vm_args for CNI & JNI invocation API 
>>- an environment variable (LIBGCJ_ARGS or some such) which sets VM  
>>parameters for any gcj binary 
>>
>
>I don't understand that last part: don't *all* gcj applications have
>to use the invocation API? I presumed that GCJ compiled (native) code
>was wrapped with some invocation caller.
>

Yep, but parameters specified on the command line to a native gcj binary 
go directly to the Java main() method, not vm_args. So if we want to be 
able to specify VM/runtime arguments as well (eg max heap size, 
classpath, debugging flags, etc), it needs to be done in an environment 
variable. It should be just a matter of having something to read the 
LIBGCJ_ARGS and pass it to _Jv_CreateJavaVM().

regards

Bryce.