Pass toolchain and user env variables to make invocation

[fmeum]

Fixes #776.

[fmeum]

@UebelAndre Would you be able to take a look at this PR? Without it, the make can't really be used safely as it does not take Bazel toolchain information into account.

[UebelAndre]

Sorry for the delay! One thing I want to clarify just to make sure I understand the change before continuing 😄

[UebelAndre]

Thank you so much for putting this together! The last thing before an approval though, would you be able to add some regression test? The rules as they are now are very prone to regressions so I think it'd be very helpful to have either a unit test in ./tests or an integration test in ./examples. Do you have a minimal repro case you'd be willing to share?

[fmeum]

Thank you so much for putting this together! The last thing before an approval though, would you be able to add some regression test? The rules as they are now are very prone to regressions so I think it'd be very helpful to have either a unit test in ./tests or an integration test in ./examples. Do you have a minimal repro case you'd be willing to share?

Could such an integration test also consist of a make project added to examples/third_party? I am making these fixes to build open-source dependencies after all, so they would be natural examples.

[UebelAndre]

It depends on the size of the project. We have some fairly large 3rd party examples currently but we feel they're worth always building since they have complex build scripts to exercise the rules and are widely used by users of rules_foreign_cc. I'd prefer a simplified example so if there were to be a regression. The iteration is fast and it's clear what the specific requirement is. Vs having to understand a larger project first to understand why a change broke support for it

[fmeum]

It depends on the size of the project. We have some fairly large 3rd party examples currently but we feel they're worth always building since they have complex build scripts to exercise the rules and are widely used by users of rules_foreign_cc. I'd prefer a simplified example so if there were to be a regression. The iteration is fast and it's clear what the specific requirement is. Vs having to understand a larger project first to understand why a change broke support for it

I rebased this onto #796 and turned make_simple into a more complete integration test verifying the presence of toolchain linkopts and rule copts. However, I am hitting a CI failure on Windows that I don't understand: cl.exe fails to create liba.o. Do you have an idea what could be wrong here? If this is a known issue with cl.exe, I could reintroduce a wrapper script that uses clang-cl instead, I just found the test to be more realistic this way.

[fmeum]

@jsharpe I reintroduced the wrapper and along the way fixed the test failure on Windows.

[fmeum]

@UebelAndre Could you take another look?

[UebelAndre]

Ah! Sorry for the delay again! Thank you so much for putting this together and awesome job on identifying the windows issue 😄