Hey Jason,
Boris and others, what about this as a suggestion. Jason seems happy
to get it out the door for this release. I suggest we do and and then
monitor 3 things before deciding whether to continue for following
releases:
1 - is anyone downloading / using P
2 - is it painful for the contributors to check they have not broken
the bindings
3 - whether Jason or someone can help when there are problems
If we have problems then we can dump it at 3.1. What do you think?
Gareth
On 20 Mar 2008, at 05:01, Jason Stewart wrote:
Hey Gareth and Boris,
I'm pretty calm about the whole thing. It's been really super being
part of the Xerces team for seven years - even if my total
contribution hasn't been all that much. I do not have any ego at stake
as to whether Xerces-P survives the chopping block.
If people want it around, then I think they should speak up.
And it would be a big help to me if others would help maintain the
bindings. But the swig bindings are beyond most Perl users (or Python
or Ruby users too). They are all C++ - and pretty involved C++.
To see if something breaks - it is now as simple as doing a make
test-suite, and we should see all tests pass. If individual tests
fail, they can be run manually to generate more output to indicate why
they failed.
Usually the problem is that some class header file changes, and some
method has a new type signature and the library won't link. When that
happens, the SWIG-generated C++ files need to be regenerated by
running swig. So usually, it is pretty simple. What has created
trouble for me in the past is upgrading to more recent versions of
SWIG to gain new features... This has caused me to rewrite the
interface definition files, and that often takes debugging time. But
if we stick with a given SWIG version, creating a new release of
xerces-p is as simple as rerunning swig and recompiling.
As I mentioned last year, I would love to see the test-suite grow to
include all the classes of the public Xerces API and all the methods.
Right now, I'm guessing the coverage is less than 50%. This may be
seen as a reason for keeping the SWIG bindings around. The Perl test
harness is VERY helpful for unit testing - but of course I'm a Perl
programmer...
Again, I'm open to suggestions.
Cheers, jas.
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
--
Gareth Reakes, CTO WE7
+44-20-7117-0809 http://www.we7.com
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
