Remove optimisation causing failures

[JonasVautherin]

Resolves #146.

This seems like an optimisation (it should be fine, though suboptimal, to sometimes run the toolchain twice) that breaks some use-cases (different people complained in #146). I therefore believe that it should be removed (and brought back properly if there is a need at some point).

Also @leetal said:

I cannot remember either why i changed that in the first place... Seems to run equally good using both the "old" solution as with the ENV solution.

Seems like it should be fine to just remove it.

[JonasVautherin]

@leetal kindly pinging here :blush:

[JonasVautherin]

I don't think this will get more attention, so I'm closing this to clean my list of open PRs.

[leetal]

I'm afraid this could break existing uses of the toolchain. I'd love to integrate this fix since it will most definitely fix more than it breaks.. I can probably add some tests that covers Conan or similar use-cases just to be sure but really the only option viable would probably be to introduce a flag to gradually roll this out.