Code review comment for lp:~thomas-voss/mir/explicit-gcc-version

> To me, coming from the mobile world, it's natural to stick with a particular,
> proven toolchain and assume others won't work until someone validates them and
> proves that they do. Toolchain bugs are usually difficult and labor intensive
> to spot and I'd rather not burn my cycles chasing them.

I agree with all that.

> Since C++11 is experimental, it makes sense to want to have control over its
> use. Though I'd have to agree with others that this is not completely solving
> the problem. But with this, we can exercise a bit better control.

My disagreement is that introducing independent control for *each C++11* project isn't obviously any improvement. But if I don't have to use this mechanism co-ordinate the changes it isn't my concern. [Dons "someone else's problem" sunglasses]

> Also, having specific toolchain dependencies could prevent potential headaches
> for phone vendors using our software.

This doesn't dictate to phone vendors. The debian/control dependencies are simply what is used for the Canonical archive.

We also build and test with clang (but we don't publish that version). We can add other compilers.

> Need fixing :
>
> As pointed out, drop the changelog mods.

Yes, that's the only blocker. Thomas agreed to change that yesterday.