https://bugs.freedesktop.org/show_bug.cgi?id=63398
--- Comment #12 from Lionel Elie Mamane <[email protected]> --- (In reply to comment #11) > As a workaround, can we not retain the original semantics by hard-coding and > categorising events that used to be 'oneway' Yes, if "someone" writes a function "wasOneWay", then sure, I'll adapt the code to call this new "wasOneWay" instead of the typedescriptor's "isOneway()" and at least we don't regress. As you write, we have to "hardcode" it. The question is how to make that "wasOneWay". I imagine we'd have to write a big function like: bool wasOneWay() { if (event == "com.sun.star.foo" || event == "com.sun.star.bar" || event == "com.sun.star.qux" || ... ) return true; else return false; } What is your suggestion? manually look at the diff of commit 90eac3e69749a9227c4b6902b1f3cef1e338c6d1? Hmm... maybe with some postprocessing of the diffstat... we get the list of changed files, strip .idl at end and "offapi" at the beginning, change "/" to "."... Now, we just have to extract which of these files are event listeners. Any idea for that? Or just put them all? Yeah, this could work. > so ... surely we just need to come up with a pseudo-oneway annotation in > raw-code for those > few methods in that module ? If we could annotate the *event* with a boolean that says: - false -> caller needs to treat this event immediately and wait for its result - true -> caller can fire the event and forget about it, or save the event for future batching with other events rather than maintaining a separate list, then that "smells" better. But I don't have an idea for annotating the event without re-introducing a new flag in IDL (which we would initially give the same value as oneway had). -- You are receiving this mail because: You are the assignee for the bug.
_______________________________________________ Libreoffice-bugs mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/libreoffice-bugs
