On Thu, Apr 03, 2003, [EMAIL PROTECTED] wrote:

> On Wed, 2 Apr 2003, Martin Andrews wrote:
> 
> > struggled with the fsl.postfix configuration. Syntax or other errors in
> > the fsl.postfix file caused postfix to dump core. Ug. [...]
> 
> The fact that changes in the fsl or errors in configuring lead to core
> dumps lets me doubt if fsl is a good approach to the problem. In general I
> had several reports of people having problems with fsl which have been
> hard to track down. E.g. a program crashes because the log file is not
> writable etc.
> 
Besides my statements below regarding "kolab disable fsl" i want you to
know that i was not able to crash postfix with syntatically or logically
wrong fsl configs. I used postfix-2.0.2-1.2.1 with fsl-1.0.8-1.2.2 for
testing. If you use these or later versions and find a configuration
which crashes please tell me about it. Make sure you rebuild
applications like postfix after an fsl upgrade - it's statically linked
and won't change until after rebuild and reinstall the application.

> Additionally because of the static nature of OpenPKG it is
> not so easy to track down.
> 
It is as easy to find out if a program uses fsl and
which version, even in a static configuration - see
http://www.ossp.org/pkg/lib/fsl/faq.html#static-fsl

> In the kolab project we therefor decided to disable fsl everywhere.
> 
We are aware of this and i have an item on by TODO list to

 - check that really all packages using syslog(3) become fsl enabled
 - every use of fsl becomes an option which defaults to "yes"

So in the near future you could easily inhibit fsl without having to
hack every spec file. This decision was primarily driven by kolab.

However, we still believe in fsl, even the current incarnation, and
continue to support it. We are currently designing a major improvment of
the underlying OSSP l2 library. This library is much more sophisticated
than syslog(3) and the tradeoff is that the more complex code could more
easily fail. This won't change because it can't.  But we're working on a
asynchronous channel which dedicates most of the logging to a different
process, so crashes won't touch the main program anymore. If this work
has been completed, fsl will inherit those improvments. It is our hope
that kolab will finally come back on the fsl track in the future. We
need the feedback of the community to improve the code.

> In general people are not used to crashes due to a non working logging
> setup. Taditionally if logging was not possible for whatever reason
> programs did continue to work.
> 
Agree, and that's what (we) the OSSP l2 team are working on.

Thanks to this thread i improved and created
- http://www.ossp.org/pkg/lib/fsl/faq.html#syslog-fsl
- http://www.ossp.org/pkg/lib/fsl/faq.html#static-fsl
Thank you for your feedback.

--
[EMAIL PROTECTED]
Development Team, Operations Northern Europe, Cable & Wireless
______________________________________________________________________
The OpenPKG Project                                    www.openpkg.org
User Communication List                      [EMAIL PROTECTED]

Reply via email to