Mir

Code review comment for lp:~raof/mir/the-least-dirty-thing

Revision history for this message
Chris Halse Rogers (raof) wrote :

> Minor observations so far:
>
> (1) 18 +Package: libmirclient-debug
> is potentially confusing with the existing convention: packagename-dbg
> so I think the dash needs removing. Something like "libmirclienttesting",
> since it's really not about "debugging" Mir at all.

It actually is a bit about debugging Mir; at least, I've used the symbols in there to debug Mir and low-level client bits like XMir.

>
> (2) It seems we're going backwards in code bloat faster than I can keep up
> with fixing it. Can we avoid this part?

Not without either (a) exposing all the libmirclient internal symbols, or (b) adding in a real extension mechanism.

Since this library isn't expected to be installed on anything but development machines I wouldn't care if it was 100MB. I don't think we should spend any significant effort caring about its size.

We'll want an actual extension mechanism at some point in the future. At that point this can be switched to using that and libmirclient-debug-extension.so won't be almost exactly the same size as libmirclient.so

>
> (3) The amount of coupling we have is a bit excessive. Or is that an existing
> issue imposed by the existing class design?; "make_mediator()"

Yeah, that's the existing class design.

« Back to merge proposal