On Thu, Sep 13, 2018 at 03:05:35PM +0200, [email protected] wrote: > I have a related question. Is the generic property multiplexing moving > forward? I remember Jake saying that he would like to have it too but that > there are some concerns about how it would fill in the current state of > Qbs (hence the current limitation to those three properties). > no progress, because it's quite some work, with little immediate pay-off (it will be necessary for a bunch of features i really want, but these are a little further off).
a precondition for generic multiplexing is fully on-demand property resolution. that is, you don't have a "resolve properties" phase any more, but instead start out with a more abstract goal (like, "construct build graph") and everything gets resolved dynamically as-needed. there are some challenges on the way here ... once that is done, the multiplexing itself is conceptually quite easy: while resolving a property, you record the axes along which its dependencies are multiplexed; during the first pass you simply use index zero on every axis, and by that you also create index 0,0,... of that new property. subsequently, you (repeatedly) resolve the property again, with the other dep indices, thus creating a fully multiplexed property in turn. there also needs to be a mechanism to de-multiplex properties, that is, to treat a particular property as an array with all elements along a particular axis rather than evaluating the expression multiple times; you'd use that in aggregator Rules like lipo. _______________________________________________ Qbs mailing list [email protected] http://lists.qt-project.org/mailman/listinfo/qbs
