I am not too sure this post landed in ARSLIST or not, so re-posting again...
Appajee On Thu, Apr 22, 2010 at 12:42 PM, Appajee Papolu <[email protected]> wrote: > Hi All, > > OK, yesterday I and David synced up (again) specifically about this topic > this time. > > As David wrote, I have recently left BMC and taking some time off > currently. > [I will forward another note about my personal decision. Surprised I have > not sent it to ARSLIST earlier]. > > Coming to the specificis of .NET API -- yes, the code repository and the > technical details are all handed off to BMC engineers. So, you should > continue to seek out information/support from BMC and show your interest > about this API. As its officially unsupported, you'll get whatever BMC > chooses to give at its discretion. > > Thanks for the kind words Rick. Yes, I had been one of the strong > proponents > of this API (and similar API wrappers) **along with few other engineers** > within the BMC engineering. So I am hopeful you will continue to get > revisions/tips regarding these pieces. > > Now, finally about the original issue that started off this thread -- "I am > getting (unknown/weird) casting error from int to object[]" or similar > lines, around active link API accesses... > > This has all the symptoms of some funny assignment or a qualification > within > action list of your offending active link. Just to give a small backdrop -- > .NET API is a managed wrapper of raw C API unmanaged code. To keep this > "plumbing" in a systematic way, the authors (well, me) chose to move > managed > data (from C# code) into an array of variants. Remember a variant can > include primitives, strings etc as well as other arrays as well! So given > any data structure, using this scheme, the .NET API converts it into a one > dimensional variant array (elements of which can be some times arrays > themselves to account for the nested nature of given structure). Well, at > this point, this array is pushed down into the unmanaged layer, where a > reverse-mapping occurs to convert this variant array into appropriate C > data > structure. As you can see, it is utterly critical that whatever is built > from C# code side, must "align" perfectly into whatever C data structure > you > have!! For the most part they do -- as we as . NET API developers coded all > of that. But, there may be cases where this is thrown off - due to a. bug > in > the .API code (mostly), b. funky situation of changed data structures or > the > values within members between API client and server versions (less likely > but possible). I feel this is a .NET API bug (which is not probably > encountered often due to the specific AL definition may not be that > common(?) perhaps). At some point, in this variant array it got, it seems > to > have an integer primitive -- where as the code was expecting an object > array > (indicative of sub-structure object likely). There is a misalignment here > between the data it has vs the variant array marshaling implementation. > > I think some action (most likely a Set/PushFields action has some thing > funny in it, specifically the action assignment expressions side) is the > culprit. May be a Macro or Ole Automation action, or perhaps SetFields > action from remote form (or, advanced mode action with sample > form/server)?? > I am just guessing here. If I can get a hand on the offending active link > definition file (as DEF or preferably XML as I can read it better), I may > be > able to guess little bit more accurately. > > Any way, if you have further questions, feel free to ping me. I would love > to help out, but be warned that I may not able to provide "support" > (investigate/debug at code level or find fix) as I do not have access to > BMC assets any more. But hopefully we can figure out a workaround or > exchange some useful info that I can share over the top of my head. > > HTH > Appajee > [email protected] > _______________________________________________________________________________ UNSUBSCRIBE or access ARSlist Archives at www.arslist.org attend wwrug10 www.wwrug.com ARSlist: "Where the Answers Are"
