Comment 1 Artyom Kuanbekov 2004-09-16 08:21:52 EDT

Created attachment 14569 [details]
64bit patch for CDT DEBUG MI 2.1 build I200409150300

This patch introduces concept of IAddress at debug MI level. It adopts chages
done in debug model by patch from bug 74056.

This patch is done against CDT 2.1 buld I200409150300. The build is available
at http://download.eclipse.org/tools/cdt/builds/2.1.0/I200409150300. 

To apply the patch do the following
1. Import org.eclipse.cdt.debug.mi.core and org.eclipse.cdt.debug.mi.ui
plug-ins 
from CDT 2.1 build I200409150300 to your workspace with sources
2. Copy cdt64bit-debug-mi.patch file to your workspace directory
3. Change dir to your workspace directory
4. Apply patch using the following command
   patch -p1 < cdt64bit-debug-mi.patch
5. Refresh exported projects in Eclipse

Comment 2 Alain Magloire 2004-09-16 17:25:42 EDT

The problem for the MI and the debug patch is , if you
look at the MI plugin for example, it was designed with three source folders
cdi/, mi/ and src/

It was done this way because we wanted to isolate the Eclipse specific parts
so we could have an mi.jar that would be independent of Eclipse and
be use elsewhere.  IAddress is define in the core plugin, so we have now
a dependency on the core Plugin.  Could not the MI plugin use BigInteger
instead ?

This will be pending after we resolve the core issues in 69908

Comment 3 Artyom Kuanbekov 2004-09-17 05:38:10 EDT

Created attachment 14603 [details]
64bit patch for CDT DEBUG MI 2.1 HEAD

I had a look into CVS and found that source directory structure is different
than in CDT 2.1 build. So I decided to produce patch against today’s HEAD
from CVS

To apply the patch
1. Import org.eclipse.cdt.debug.mi.core plug-in from CVS to your workspace with
sources
2. Apply patch using Team->Apply Patch context menu (Ignore path segments
should be 2)

NOTE: Plug-in org.eclipse.cdt.debug.mi.ui do not need patching

Comment 4 Artyom Kuanbekov 2004-09-17 05:51:22 EDT

Reply for comment #2

IAddress is used only in two MI events
1. MIMemoryChangedEvent
2. MIMemoryCreatedEvent

To get rid of dependency on CDT CORE we can
1. Use String instead of IAddress (Simple, conversion functions are already 
present)
2. Use BigInteger instead of IAddress (Conversion functions should be added)

Could you please also clarify how MI Events may be used outside Eclipse? As far 
as I understand these events are fired by MISession but this class is not 
included into mi source folder? May be we can move MI Events into src folder?

Comment 5 Alain Magloire 2004-09-17 14:31:52 EDT

> To get rid of dependency on CDT CORE we can
> 1. Use String instead of IAddress (Simple, conversion functions are already 
present)
> 2. Use BigInteger instead of IAddress (Conversion functions should be added)

Sounds good to me, first let me deal with your patch we clean this
later.


> Could you please also clarify how MI Events may be used outside Eclipse? 

All the MIEvents are in the mi/ src folder a few classes are still
in src/ MIsession, MIInferior etc .. because of there dependencies
on core via PTY and spawner.

> As far as I understand these events are fired by MISession but this class
> is not included into mi source folder? May be we can move MI Events 
> into src folder?

The idea is that in the future we would like to use the MI code parsing
outside of eclipse and distributed them as as a separate jar.  We do
this for example in the Core plugin, the parser AST is in separate jar
cdtparser.java.  There is not Eclipse class dependencies in that source folder.

At this moment, it is not possible because of the dependency
for PTY and Spawner class in core, but we are looking at ways to get rid of
that for the future.

Comment 6 Artyom Kuanbekov 2004-09-29 07:12:34 EDT

Just checked out latest CDT 2.1 build and noticed that CDI and MI levels use 
BigIntegers for addresses now. I think it is worth to add IAddressFactory.
createAddress(BigInteger) so performance does not degrade from extra BigInteger-
>String and String->IAddress conversions.

Thanks.

Comment 7 David Inglis 2004-09-29 09:05:47 EDT

sound like a good idea, I'll add it.

Comment 8 Nobody - feel free to take it 2004-09-29 14:24:43 EDT

IAddress should also provide a method that returns the size of address (in 
bytes or bits).

Comment 9 David Inglis 2004-10-01 10:54:30 EDT

>IAddress should also provide a method that returns the size of address (in 
>bytes or bits).

sounds reasonable, I'll add getSize() with return being size in bytes

Comment 10 Nobody - feel free to take it 2004-10-01 11:03:37 EDT

Fantastic! Please, let me know when you are done. I have to apply these changes 
to the Memory view.

Comment 11 Alain Magloire 2004-11-03 10:56:01 EST

Patch schedule for 2.1

Comment 12 Alain Magloire 2004-11-03 10:56:42 EST

Patch and changes in the head for CDT-2.1

Please address other issues in a new PR