Allow the taken flag to be optimized out

[jonas-schievink]

Fixes #242

Verified manually that it still prevents linking in multiple versions.

[rust-highfive]

r? @thejpster

(rust_highfive has picked a reviewer for you, use r? to override)

[adamgreig]

I'm happy with this change to get us lower-cost for the time being, as far as I can tell it preserves the same guarantees while allowing some cases (especially e.g. RTIC) to eliminate the superfluous bool.

However I do note that #239 still has the same issue that both an operating system and an application/driver may need to use cortex-m, but be linked totally separately and only combined at runtime, in which case this "safe" interface is not; perhaps ultimately cortex-m should lose take() and the singleton safety be provided by cortex-m-rt or other runtime.

bors merge

[therealprof]

LGTM

[bors]

Build succeeded:

• continuous-integration/travis-ci/push

[therealprof]

However I do note that #239 still has the same issue that both an operating system and an application/driver may need to use cortex-m, but be linked totally separately and only combined at runtime, in which case this "safe" interface is not; perhaps ultimately cortex-m should lose take() and the singleton safety be provided by cortex-m-rt or other runtime.

I agree.