Matthias Kramm

The pdf might be using a font that gfx2gfx can't find. Try running "pdffonts" (part of xpdf) to find out what font it is, and then add it to /usr/local/share/swftools/fonts...

I don't see why not. I'll happily accept pull requests implementing this. It looks like Speex has an API, so this should hopefully be pretty straightforward.

Yes, it would be good to fix this. I'll take a PR.

Arthur: I'm afraid it's broken in your branch as well - both use f_back to link cells to their function, instead of using the closure parameter of pyobj.Function. The f_back...

If mypy deals better with `Any` returns instead of `Union` returns, would it be possible to have mypy interpret any `-> Union[...]` in typeshed as `-> Any`? (Btw. pytype does...

Yeah, this one's up for grabs, as far as I can tell. Glad to hear you'll take it on, Ofek! Happy hacking. :) Let me know if I can help...

I agree, you could (re)use DenseAnalysis.cpp, make it bidirectional. What pattern were you thinking of using for determining the direction? Overloads? "if (this->forward) {...} else {...}"? Some kind of visitor...

Not sure the solver should be direction-aware. Direction seems more a property of the callgraph / lattice, not of the solver. (Then again, the solver comments say the analysis *starts*...

I've been wondering whether we need the concept of a "direction boolean" (or overload) at all. Suppose our lattice elements support both join() and meet(), and both setEntryState() and setExitState()....

> can you elaborate more on what you mean with a concrete example? Currently the dense analysis implements `join` as: ``` void join(Lattice *lhs, const Lattice &rhs) { propagateIfChanged(lhs, lhs->join(rhs));...