On 4/12/07, Venkata Krishnan <[EMAIL PROTECTED]> wrote:
Hi Simon, Thanks for considering on expanding the 'problem model'. Infact I was wondering if every SCA Object such as a Composite, Component, Reference, .... has a bunch of 'error contexts' encapsulated within it. So if something is wrong with a Reference defn, you simply add the problematic 'reference defn. instance' (as it is being done now) and also the Error Context say something as 'MULTIPLICY_TO_TARGET_MISMATCH'. The for each of these 'Error Contexts' the SCA Object also encapsulates what information would be relevant to output as Error Information. So if you'd as for an error message to the Reference Object given the context 'MULTIPLICY_TO_TARGET_MISMATCH' it would provide you the multiplicity and target settings. So we simply stack the problematic SCA Objects and simply ask for Error Information at the end. Not sure it this would complicate things... it was just a thought... maybe there are better options. - Venkat On 4/12/07, Simon Laws <[EMAIL PROTECTED]> wrote: > > On 4/12/07, Jean-Sebastien Delfino <[EMAIL PROTECTED]> wrote: > > > > Jean-Sebastien Delfino wrote: > > > [snip] > > > Simon Laws wrote: > > >> > > >> Composite configuration problem: > > >> [EMAIL PROTECTED] > > >> > > >> I've not looked into this one specifically but it doesn't stop the > test > > >> passing. I do get more of these problem reports in the composite-impl > > >> test > > >> that I'm playing with and it is a real problem in that case. Still > > >> looking > > >> at whats going on. > > >> > > >> Regards > > >> > > >> Simon > > >> > > > Simon, > > > > > > This output is produced by temporary test code that I added to > > > CompositeUtil to help track problems in most of .composite files, > > > which are not all following the SCA assembly XML spec 1.0. In > > particular: > > > - declare a targetNamespace > > > - use qnames to name referenced composites > > > - name included composites in the content of the <include> element > > > instead of a "name" attribute > > > - use "target" attributes on references instead of naming the target > > > in the element content > > > - use "promote" attributes on composite services and references > > > > > > So before testing, you need to make sure that the .composite files are > > > correct. These print statements should help you detect that there are > > > problems (even if they just dump the model objects at the moment). > > > Then to debug you can set a breakpoint on the line that does the > > > print, and figure what 's wrong in the test case from there. > > > > > Simon, > > > > I just committed a change to print more debug info when encountering a > > composite configuration problem. Hope this helps. > > > > -- > > Jean-Sebastien > > > > > > --------------------------------------------------------------------- > > To unsubscribe, e-mail: [EMAIL PROTECTED] > > For additional commands, e-mail: [EMAIL PROTECTED] > > > > Great, thanks Sebastien.Thanks for doing that. I gave the new code a > spin > and the extra info is interesting. Looking at the CompositeUtil code is > looks like it is always comparing the component definition from the SCDL > with the implementation. Do you have any objection if I go in and add some > more info to the problem model to record and report the exact nature of > the > problem found? > > Simon > > Simon >
I like the idea that when capturing the state of the object we capture the error state as well. A question though. Were you thinking of doing this with the model objects in the assembly or, when you say "SCA Object", do you mean passing the info through to the runtime objects some how. If the latter not sure how you would do this as I assume they are not around until after the build stage. If the former then that sounds OK to me. It would be a more detailed version of the "unresolved" flag, I.e. I tried to resolve it but found errors. As an aside why is that flag called "unresolved"? Shouldn't it be called "resolved". I struggle with the Unresolved=false double negative. (not a major point - feel free to ignore me:-) Regards Simon
