On Mon, Jan 26, 2004 at 10:18:55PM +0100, Cédric Chabanois wrote:
> >- Fix the keyword test (ignore the authors name since that depends
> > on the user who runs the tests).
> >
> It should work as it because in keyword test the username is set :
> <svn username="cedric" javahl="${javahl}">
>
> It should work (didn't it ?)
No, the test result was "$Author: digulla $" -> failure. Maybe this
only works with javahl?
> >I'm wondering if you should drop the support for javahl, (at least for
> >now):
> >
> >1. It's not a stable/really supported interface.
> It is a supported interface, at least on windows.
> It works better than command line interface on windows.
> And it will be a supported interface on Linux too when javahl will be
> integrated in the subversion build process.
> McClain Looney also talked about a statically linked version of javahl
> too ...
For some reason, the direct interfaces (Python, Java, whatever) don't get
the same attention than the rest of subversion. For example, the example
scripts are not updated when the core changes. Also, the language definition
files (SWIG) don't get the attention they need which means that for 4 out
of 5 subversion releases, you can't compile them.
Plus, compiling this stuff usually depends on a specific version of the
language, etc.etc.etc.
In the end, when you want to support javahl, you must either supply a working
DLL for all platforms on which you can have eclipse (that's seven, when I last
looked: Windows, Linux, Solaris, QNX, AIX, HP-UX and MacOSX) *or* you must
state that subclipse is Windows only. As it is currently, it's very irritating
because you say that you support Linux but in fact, you don't (The current
state of the Linux version basically says "Windows only" despite your efforts).
> >2. You must create the dll for every platform (linux, windows, mac,
> >solaris,
> > HP-UX, AIX, etc. etc.)
> Yes but it is more efficient than command line interface.
Let me rephrase that: You refuse to create something which works
everywhere because it works more efficiently in one single place.
That's not really efficient. :-)
> >3. The dll is very tightly integrated into the executable while the
> > commandline version doesn't have to be updated for every version
> > of subversion.
> >
> The commandline version needs to be updated too. At least Subversion
> 0.35.1 broke it
Let me rephrase that, too: The command line cient broke 3-4 times in
the last 30 releases. The javahl-based version broke 30 times in the
last 30 releases.
Which one is more efficient? :-)
> We will try to support both interfaces (at least on windows and linux
> for javahl).
Well, I fear that's not very efficient as long as subversion changes its
API so quickly. I would prefer a 100% working command line interface
until subversion stabilizes and all official releases include javahl.
--
Aaron "Optimizer" Digulla a.k.a. Philmann Dark
"It's not the universe that's limited, it's our imagination.
Follow me and I'll show you something beyond the limits."
http://www.philmann-dark.de/
Received on Wed Jan 28 06:12:33 2004