Re: native java libraries

2005-03-29 Thread Jerry Haltom
I was responding to this: >>>> So here is my proposal: Since the use of native java libraries is a feature of GCJ/GIJ only,the handling of those libraries should be left to those. GIJ should first check if there is native version of a java library. If not, it should compile one th

Re: native java libraries

2005-03-29 Thread Michael Koch
On Tue, Mar 29, 2005 at 08:33:47AM -0600, Jerry Haltom wrote: > There is no good way to make a global cache at runtime other than a SUID > binary, which is not happening. A per user cache is also not practical. Package installation is done as root so no SUID bit would need to be set. And it doenst

Re: native java libraries

2005-03-29 Thread Michael Koch
tallation, the usual approach > of a python interpreter is to compile such files on demand. I think, such an > approach would be a good > idea for native java libraries. > > So here is my proposal: > Since the use of native java libraries is a feature of GCJ/GIJ only, the > handli

Re: native java libraries

2005-03-29 Thread Michael Koch
On Tue, Mar 29, 2005 at 01:24:26PM +0200, David N. Welton wrote: > "Michael Koch" <[EMAIL PROTECTED]> writes: > > > As discussed earlier this is a huge task with no clear benefit. We > > have to solve some issues before we can even consider doing it: We > > need GCJ 4.0 in unstable and someone nee

Re: native java libraries

2005-03-29 Thread Jerry Haltom
for implementation details... > > > Though in Debian the Python byte-code-files seem to be compiled on > installation, the usual approach of a python interpreter is to compile > such files on demand. I think, such an approach would be a good idea for > native java libraries. &g

Re: native java libraries

2005-03-29 Thread David N. Welton
"Michael Koch" <[EMAIL PROTECTED]> writes: > As discussed earlier this is a huge task with no clear benefit. We > have to solve some issues before we can even consider doing it: We > need GCJ 4.0 in unstable and someone needs to make performance tests > to proof this gains anything at all. For mor

Re: native java libraries

2005-03-29 Thread spam2005
ompiled on installation, the usual approach of a python interpreter is to compile such files on demand. I think, such an approach would be a good idea for native java libraries. So here is my proposal: Since the use of native java libraries is a feature of GCJ/GIJ only, the handling of thos

Re: native java libraries

2005-03-28 Thread Michael Koch
On Tue, Mar 29, 2005 at 01:37:58AM +0200, Philipp Hug wrote: > Why don't we optionally compile the jars when the package is installed on the > users machine? > This would save a lot of archive space. > it would take longer to install java packages if you have this option > enabled, > but as it i

native java libraries

2005-03-28 Thread Philipp Hug
Why don't we optionally compile the jars when the package is installed on the users machine? This would save a lot of archive space. it would take longer to install java packages if you have this option enabled, but as it is up to the user to enable precompilation, this shouldn't be a problem.