Improve include directories support

[hlopko]

Project to improve expressiveness of includes. We need non transitive includes, proper support in the crosstool, and we need to stop parsing copts to detect all includes for header validation.

Problems:

• 'includes' produce '-isystem'
• Using strip_include_prefix changes -isystem into -I
• Non transitive includes are only possible using copts, bazel parses copts to collect all include directories for include validation
• The order in which includes appear on the command line is unclear Order of include path entries on the command line is unspecified #2143
• We don't use nested sets for includes/defines.
• It's not clear if the current order is compatible with layering check
• There is no way of specifying C-specific include directories C++ code fails to build with clang 4.0 #3977

Related issues:

• Failed to build using -isystem #3206
• includes and defines should not propagate further than direct dependents #3271
• cc_library rule should also be able to strip_srcs_prefix #3212
• Failed to build using -isystem #3206
• Make specification of 'cc_library' header dependencies platform-independent. #2172
• Order of include path entries on the command line is unspecified #2143
• Accept label_string_dict for includes #2611
• Always pass -D__BAZEL__ to C compiler #2160

[hlopko]

Relevant discussion: #6343

[scentini]

This will not make it into 1.0.

[elklein]

Any news on the status of this issue? I've hit up against it. I'm able to hack around it be using a pre-existing wrapper that we have around gcc (so that we can dispatch work to either gcc or nvcc) to hack certain -I flags over to -isystem instead, but that's clearly not a very nice solution. Would love to be able to use a proper system to designate which includes should be -I vs. -isystem. Thanks!

[sgowroji]

Hi there! We're doing a clean up of old issues and will be closing this one. Please reopen if you’d like to discuss anything further. We’ll respond as soon as we have the bandwidth/resources to do so.