Re: [gomd-devel] <DAEMON>: working on the SCX

[Gian Paolo Ghilardi]

Hi all.

----- Original Message -----
From: "Matthias Rechenburg" <address@hidden>
To: "gomd developers mailing list" <address@hidden>; "Gian Paolo
Ghilardi" <address@hidden>
Sent: Thursday, September 11, 2003 11:38 AM
Subject: Re: [gomd-devel] <DAEMON>: working on the SCX


> Hey JP,
>
> On Donnerstag 11 September 2003 02:46, Gian Paolo Ghilardi wrote:
> > Hi all.
> >
> > I'm working on the SCX feature.
> > Actually I can run a program in a separate process and, finally, I get
the
> > child output from the main process.
> > I've still a problem with the execXX() family of calls. Some programs
like
> > "ls" work fine but others like "find", not!
>
> no idea yet. how can i test it ? is it already in the cvs ?
Sorry. I'm working on.
I've still the problem with execXX() family.
Moreover I've to link the code with ACL facility and the new comamnd check
mechanism (yet coded).

> .... btw: i changed my mind and now i also think that it would be
> nice to support long-running commands if possible  ;))
The idea is to support every kind of process.
We need a timeout for the process => after some time the process must be
killed for security reasons.
The timeout will be placed in constants.h.

>
> >
> > BTW: these day I'm busy (studying Economics for an exam (19 Sept)) but,
> > when I can, I continue with gomd devel.
>
> have a good exam
Thanks... ;)

>
> > I want to release the beta ASAP.
>
> ok :)
>
> >
> > Byez.
> >
> > Johnny: I like pthreads. Commoncpp provides a Thread wrapper class
(based
> > on pthread, I suppose) but it's not so handy as I'd like it was.
> > ATM the code uses standard fork() but I'd like to use threads.
>
> mhmm, ok, which advantages are giving pthreads above forks ?
> ..... forked procs may even migrate ;) pthreads don't
Uhm...
Procs using FSU-Pthreads can be migrated (these threads are in user-space).
Linux pthreads implementation is in user-space but threads can't be migrated
(becaus ethey are too bound to system).
LinuxThreads (kernel threads, created by __clone()) can me migrated via
misgshm patch.

ATM the code uses procs but can be easily moved to support threads (maybe
via the commoncpp wrapper class).

<rejected>