Mikhail Loenko wrote:
Hi George
if you restructured the stuff on your computer, could you submit a patch?
Thanks,
Mikhail
Hi Mikhail,
Yes, it has always been my hope to do so. Given the amount of change
involved, I just wanted to run things by the broader community before
proceeding to open a JIRA on the matter. Your feedback has been very
helpful and I will work towards getting the patch submitted today.
Best regards,
George
IBM UK
2006/2/27, Mikhail Loenko <[EMAIL PROTECTED]>:
2006/2/27, George Harley <[EMAIL PROTECTED]>:
Mikhail Loenko wrote:
Hi George
actually the native code we have in security should work on both
IA32 and IPF
So, it seems that with your suggestion we will have to have
two copies of that code. Please correct me if I'm wrong
What is about the following str:
+-win/
| |
| +--IA32/
| |
| +--IPF/
| |
| +-- common1.cpp
| |
| +-- common2.cpp
| |
...
Thanks,
Mikhail
Looks good to me. And it's the same story under the "linux" folder ?
Exactly.
Thanks,
Mikhail
Best regards,
George
IBM UK
2006/2/27, George Harley <[EMAIL PROTECTED]>:
Jean-frederic Clere wrote:
Mikhail Loenko wrote:
Hi George,
why e.g. 'win.IA32' not just 'win'?
Because there will be a posix.apr that will do the portable part ;-)
Correct me if I am wrong.
...er...well, I'm not sure that it was foremost in my thoughts when I
was working through the layout changes :-)
I simply wanted to have a way of differentiating between code written
for 32-bit and 64-bit Windows on Intel architecture. I am assuming that
the Windows code there today is for 32 bit. I did wonder about splitting
those directory names up so that instead of a folder called "win.IA32"
we had a "win" folder with a "IA32" sub-folder (and likewise for Linux).
i.e.
java
|
+-common
|
+-win
| |
| \---IA32
|
+-linux
| |
| \---IA32
|
...
The above approach leaves the way open for other variants (e.g. 64-bit
code) to be added in new sub-folders beneath "win" and "linux". In the
end I opted for consistency with the "win.IA32" and "linux.IA32" names
that are currently being used under the trunk/native-src folder in SVN.
Best regards,
George
IBM UK
Cheers
Jean-Frederic
Thanks,
Mikhail
2006/2/24, George Harley <[EMAIL PROTECTED]>:
Hi,
Redrawing the proposed layout as it didn't render quite correctly
for me
when I read over the sent note (sigh).
<SECURITY ROOT>
|
|
+---src
| |
| +---main
| | |
| | +---java
| | | |
| | | +---common
| | | |
| | | +---linux.IA32
| | | |
| | | \---win.IA32
| | |
| | +---native
| | | |
| | | +---linux.IA32
| | | |
| | | \---win.IA32
| | |
| | \---resources
| | |
| | \---common
| |
| +---test
| |
| +---java
| |
| +---common
| |
| +---linux.IA32
| |
| \---win.IA32
|
+---doc
| |
| \---images
|
+---make
| |
| \---native
| |
| +---linux
| |
| \---windows
|
+---META-INF
Best regards,
George
IBM UK
George Harley wrote:
Hi,
Earlier on today I spent some time following the instructions for
developing Harmony Java code inside Eclipse [1]. After experimenting
with archive, luni and nio I decided to check out modules/security and
found that, in its current form, it can't be brought into an Eclipse
workspace and used like the other modules. One obvious difference is
that it doesn't have any Eclipse project metadata in there (e.g.
.project and .classpath files). After adding these in (in my private
workspace), I began to look at other differences between security and
its peer modules in particular the difference in source layouts.
Recalling some ideas for layouts that have been kicked around this
list,
I started to move things around a little to try and make things a
little
more uniform with respect to those peer modules. Things were made more
interesting by virtue of the following security module distinctions :
* it has platform-specific Java code
* it contains native code for both Windows and Linux
Eventually I arrived at a structure that is more attuned to the other
modules in the repository. As a bonus, Eclipse pointed out several
missing import entries from the META-INF/MANIFEST.MF file - including
one that cannot presently be satisfied with what is in the Harmony
repository (org.apache.harmony.security.test.SecurityTest wants to
import java.util.logging.LoggingPermission which doesn't exist in the
repository - although an implementation has been contributed [2]).
In addition to moving source around, I also made the necessary
tweaks to
the Ant scripts contained in the security module plus the "top level"
Java build file trunk/make/build-java.xml so the Ant builds still work
as before.
Keeping my fingers crossed that the next bit of this note renders
alright in your mail client, here is the modules/security structure
that
I ended up with (minus all of the package sub-folders for clarity) :
<SECURITY ROOT>
|
|
+---src
| |
| +---main
| | |
| | +---java
| | | |
| | | +---common
| | | |
| | | +---linux.IA32
| | | |
| | | \---win.IA32
| | | | | +---native
| | | |
| | | +---linux.IA32
| | | |
| | | \---win.IA32
| | |
| | \---resources
| | |
| | \---common
| | | +---test
| |
| +---java
| |
| +---common
| |
| +---linux.IA32
| |
| \---win.IA32
|
+---doc
| |
| \---images
|
+---make
| |
| \---native
| |
| +---linux
| |
| \---windows
|
+---META-INF
All of the leaf folders under "src" have been declared to Eclipse as
source folders (i.e. I have 9 source folders called
"src/main/java/common", "src/main/resources/common",
"src/main/native/linux.IA32", "src/test/java/common" and so on...).
I would be really keen to hear what people think of this prototype
re-structuring. It would be great if we could make the security module
as simple to work with inside Eclipse as the other modules are.
Best regards,
George
IBM UK
[1]
http://incubator.apache.org/harmony/subcomponents/classlibrary/dev_eclipse.html
[2] http://issues.apache.org/jira/browse/HARMONY-88
