Ben Boeckel

Now just to get the tests to pass and figure out where to send patches…

Yeah, I'll try that. Sourceforge is being quite prickly and 403'ing any attempts to recover a password.

https://github.com/xiph/flac/pull/1

Yes, the CMake package generated by libtiff upstream does not provide `TIFF_LIBRARIES`, so nothing ends up being passed to the `try_compile` for linking.in order to pass. The check needs reworked...

Then again, I'm not sure if that allows the "triangle setup". For example, `origin` for me is *always* the "main" repo (I never got the `upstream` for that and `origin`...

It looks like GDC's `-M` flags don't work :( . I'm getting empty depends lists in CMake, so something needs to be done for it.

Look at `cmcldeps`. That is probably what we need for Ninja. It'd be nice if make would then use that mechanism. Without looking, I think it takes some arguments which...

One thing to do could be to trace the public/private imports and not print private imports as dependencies. Scanning per-target is probably faster, but I don't think ninja is set...

> So, if you have a chain of modules a, b, c, and d, each one importing the next, generating the deps for each module individually ends up with duplicated...

On Fri, May 09, 2014 at 12:25:08 -0700, trentforkert wrote: > One downside to my current implementation is that it will parse > `/* import std.algorithm; */` or similar without...