Let's not waste too much time with the in-process worker, it may be
cool but I think it's more important to get the socket channel
and the core stuff tested so people can start using jk2.
There are many interesting issues with running in process, that's
one of the reasons KDE and Mozilla seem to choose out-of-process.
We can keep it as 'experimental', and play with it more after the
first release.
The JNI methods are more important - i.e. Java + libjkjni.so,
as it allows java to do use more efficient transports and use
OS-specific features that are not available. The unix channel
and other out-of-process mechanisms we can add will probably
bring a lot of improvements.
Costin
On Wed, 5 Jun 2002, GOMEZ Henri wrote:
> >> But I've got problem with putenv/getenv calls, which didn't
> >> appears to works (did there is a problem if we want to
> >> putenv() from inside a process ?).
> >
> >Yes, most of the times it is. I don't know exactly why, it may
> >depend on the VM ( because I got it to work with some VMs ).
>
> Not only with VM, you does a putenv() at some part of the code,
> a following getenv() show the var, but when you redo getenv()
> later (ie in replaceProperties), you get null (and it's the same
> pid and threadid...)
>
> >Setting the env before starting the process seems the best solution.
> >( I suspect LD_LIBRARY_PATH is used by ld.so before the process
> >starts, and changes are not picked up ).
>
> LD_LIBRARY_PATH is mandatory before app launch
>
> >Or it may be something else - but even if we get it to work on
> >linux, there are many other platforms where it may brake.
> >
> >> >
> >> > void init() {
> >> > apr.setUser();
> >> > }
> >> >}
> >> >
> >> >It's more flexible.
> >>
> >> Did you plan to add setUser calls ?
> >
> >
> >Eventually. I'm a bit tired right now, I spent 2 days building and
> >finding dependencies. If you have some time...
> >
> >
> >> >I think late - to make sure logs and other resources can be
> >acquired (
> >> >port 80, etc ).
> >>
> >> Hum, so jni/java could get root protected resources ?
> >> I didn't understand well here, since jni/vm is launched by a child
> >> and apache2 child are no more root, (only main process stay root)
> >> so they couldn't get access to resource (and that's why I've got
> >> problem with shm access, which was created by the main process which
> >> is root).
> >
> >That's a tricky one. Of course, the quick workaround is to change
> >permissions. I'm curious what would happen if we start the VM
> >in _parent_,
> >before forking. We can then get access to root-protected
> >resources, fork,
> >than hope it'll work.
> >
> >I never tried fork from a JVM. Interesting...
> >
> >But except in-process, I think tomcat should first acquire the
> >resources
> >it needs and then call setUser().
>
> I still have to figure when parent and when childs call
> modules funcs
>
> --
> To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]>
> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
>
>
--
To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]>
For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
