[llvm] [CI] Handle Runtime Build Failures Correctly in generate_test_report (PR #155730)

[David Spickett via llvm-commits]

================
@@ -27,6 +27,13 @@ def _parse_ninja_log(ninja_log: list[str]) -> list[tuple[str, str]]:
             # We hit the end of the log without finding a build failure, go to
             # the next log.
             return failures
+        # If we are doing a build with LLVM_ENABLE_RUNTIMES, we can have nested
+        # ninja invocations. The sub-ninja will print that a subcommand failed,
+        # and then the outer ninja will list the command that failed. We should
+        # ignore the outer failure.
+        if ninja_log[index - 1].startswith("ninja: build stopped:"):
----------------
DavidSpickett wrote:

Was going to say why not check that the two commands are the same for a bit more certainty that you're in this specific situation, but they won't be, right?

The outer ninja's command will be something like "move to runtimes directory and invoke ninja" and the inner one is "build this runtime source file".

https://github.com/llvm/llvm-project/pull/155730