[all-commits] [llvm/llvm-project] 5369d8: [libc++] Remove support for .run.fail.cpp tests

[Louis Dionne via All-commits]

  Branch: refs/heads/master
  Home:   https://github.com/llvm/llvm-project
  Commit: 5369d8cca7e6443703dd4dbaf253012e4c5bb400
      https://github.com/llvm/llvm-project/commit/5369d8cca7e6443703dd4dbaf253012e4c5bb400
  Author: Louis Dionne <ldionne at apple.com>
  Date:   2020-11-03 (Tue, 03 Nov 2020)

  Changed paths:
    R libcxx/test/libcxx/selftest/run.fail.cpp/compile-error.run.fail.cpp
    R libcxx/test/libcxx/selftest/run.fail.cpp/link-error.run.fail.cpp
    R libcxx/test/libcxx/selftest/run.fail.cpp/run-error.run.fail.cpp
    R libcxx/test/libcxx/selftest/run.fail.cpp/run-success.run.fail.cpp
    M libcxx/utils/libcxx/test/format.py

  Log Message:
  -----------
  [libc++] Remove support for .run.fail.cpp tests

Unfortunately, executing these tests correctly on platforms that do not
support a shell is very challenging. Since the executor can't just negate
the result of the command, we'd have to ship a portable program capable
of running the actual test executable, and negating its result.

Doing this portably is challenging. Since we do not currently have strong
use cases for tests that fail at runtime (we effectively have no tests
using that capability right now), it is difficult to justify making them
work portably. Instead, it makes more sense to remove this feature until
we can implement it properly (i.e. without requiring shell support).