Re: Reproting (Was Re: Orocos meeting: hotel reservations)
| Newsgroups | gmane.science.robotics.orocos.user |
|---|---|
| Message-ID | <[email protected]> |
Looking forward to the discussion of the topic. Cezary ----- Original Message ----- From: <[email protected]> To: <[email protected]> Sent: Wednesday, October 23, 2002 10:47 AM Subject: Reproting (Was Re: [Orocos] Orocos meeting: hotel reservations) > On Wed, 23 Oct 2002 [email protected] wrote: > > > You mentioned that reporting is a problem. I agree to that. Although MRROC++ > > has not solved the problem completly it is not at all bad in that respect. > > The system is divided into separate processes. Each process can be viewed as > > a finate state automaton. Within each "state" of the automaton some actions > > are performed that can lead to abnormal situations that have to be reported. > > In that case an exception is thrown and subsequently caught at the uppermost > > level (the automaton level). From there the report is sent to the System > > Response Process (SRP) that visualises the message on the screen. Standard > > status messages can also be sent to the SRP, but here you can do it from any > > level of the process structure. After reportiing an error the automaton is > > being reset to a state that it can handle new commands (dialogue with other > > processes). It should be noted that after throwing an exception it can be > > caught at some intermediate level and handled adequately. In that case the > > exception does not reach the uppermost level, thus the remedied problems are > > not reported - unless necessary. The programmer is free to either use the > > general method or to override it with local remedies. I think that something > > similar could be used within the OROCOS project. We can talk about that at > > the meeting. > > I'm looking forward to it! > We still remember the emphasis you've put on exception handling during > previous Orocos meetings, and we have not forgotten about it :-) So, > Peter Soetens will present our current design, whose functionality is > not too different from what you describe. The design problem, however, > is with making sure that _everything_ in the whole framework uses a > coherent reporting interface. > > Herman > > _______________________________________________ > Orocos mailing list > [email protected] > http://mail.mech.kuleuven.ac.be/mailman/listinfo/orocos