[lldb] Add ability to detect darwin host linker version to xfail tests

[bulbazord]

When Apple released its new linker, it had a subtle bug that caused LLDB's TLS tests to fail. Unfortunately this means that TLS tests are not going to work on machines that have affected versions of the linker, so we should annotate the tests so that they only work when we are confident the linker has the required fix.

I'm not completely satisfied with this implementation. That being said, I believe that adding suport for linker versions in general is a non-trivial change that would require far more thought. There are a few challenges involved:

• LLDB's testing infra takes an argument to change the compiler, but there's no way to switch out the linker.
• There's no standard way to ask a compiler what linker it will use.
• There's no standard way to ask a linker what its version is. Many platforms have the same name for their linker (ld).
• Some platforms automatically switch out the linker underneath you. We do this for Windows tests (where we use LLD no matter what).

Given that this is affecting the tests on our CI, I think this is an acceptable solution in the interim.

[llvmbot]

@llvm/pr-subscribers-lldb

Author: Alex Langford (bulbazord)

Changes

When Apple released its new linker, it had a subtle bug that caused LLDB's TLS tests to fail. Unfortunately this means that TLS tests are not going to work on machines that have affected versions of the linker, so we should annotate the tests so that they only work when we are confident the linker has the required fix.

I'm not completely satisfied with this implementation. That being said, I believe that adding suport for linker versions in general is a non-trivial change that would require far more thought. There are a few challenges involved:

• LLDB's testing infra takes an argument to change the compiler, but there's no way to switch out the linker.
• There's no standard way to ask a compiler what linker it will use.
• There's no standard way to ask a linker what its version is. Many platforms have the same name for their linker (ld).
• Some platforms automatically switch out the linker underneath you. We do this for Windows tests (where we use LLD no matter what).

Given that this is affecting the tests on our CI, I think this is an acceptable solution in the interim.

---

Full diff: https://github.com/llvm/llvm-project/pull/83941.diff

2 Files Affected:

• (modified) lldb/packages/Python/lldbsuite/test/lldbplatformutil.py (+40)
• (modified) lldb/test/API/lang/c/tls_globals/TestTlsGlobals.py (+1)

diff --git a/lldb/packages/Python/lldbsuite/test/lldbplatformutil.py b/lldb/packages/Python/lldbsuite/test/lldbplatformutil.py
index c4d063d3cc77ef..acb1a801e7f638 100644
--- a/lldb/packages/Python/lldbsuite/test/lldbplatformutil.py
+++ b/lldb/packages/Python/lldbsuite/test/lldbplatformutil.py
@@ -3,6 +3,7 @@
 
 # System modules
 import itertools
+import json
 import re
 import subprocess
 import sys
@@ -16,6 +17,7 @@
 from . import lldbtest_config
 import lldbsuite.test.lldbplatform as lldbplatform
 from lldbsuite.test.builders import get_builder
+from lldbsuite.test.lldbutil import is_exe
 
 
 def check_first_register_readable(test_case):
@@ -333,3 +335,41 @@ def expectedCompiler(compilers):
             return True
 
     return False
+
+
+# This is a helper function to determine if a specific version of Xcode's linker
+# contains a TLS bug. We want to skip TLS tests if they contain this bug, but
+# adding a linker/linker_version conditions to a decorator is challenging due to
+# the number of ways linkers can enter the build process.
+def darwinLinkerHasTLSBug():
+    """Returns true iff a test is running on a darwin platform and the host linker is between versions 1000 and 1109."""
+    darwin_platforms = lldbplatform.translate(lldbplatform.darwin_all)
+    if getPlatform() not in darwin_platforms:
+        return False
+
+    linker_path = (
+        subprocess.check_output(["xcrun", "--find", "ld"]).rstrip().decode("utf-8")
+    )
+    if not is_exe(linker_path):
+        return False
+
+    raw_linker_info = (
+        subprocess.check_output([linker_path, "-version_details"])
+        .rstrip()
+        .decode("utf-8")
+    )
+    parsed_linker_info = json.loads(raw_linker_info)
+    if "version" not in parsed_linker_info:
+        return False
+
+    raw_version = parsed_linker_info["version"]
+    version = None
+    try:
+        version = int(raw_version)
+    except:
+        return False
+
+    if version is None:
+        return False
+
+    return 1000 <= version and version <= 1109
diff --git a/lldb/test/API/lang/c/tls_globals/TestTlsGlobals.py b/lldb/test/API/lang/c/tls_globals/TestTlsGlobals.py
index dfe29b451df0a6..526e49b68162f3 100644
--- a/lldb/test/API/lang/c/tls_globals/TestTlsGlobals.py
+++ b/lldb/test/API/lang/c/tls_globals/TestTlsGlobals.py
@@ -40,6 +40,7 @@ def setUp(self):
     @skipIfWindows
     @skipIf(oslist=["linux"], archs=["arm", "aarch64"])
     @skipIf(oslist=no_match([lldbplatformutil.getDarwinOSTriples(), "linux"]))
+    @expectedFailureIf(lldbplatformutil.darwinLinkerHasTLSBug())
     def test(self):
         """Test thread-local storage."""
         self.build()

[kastiglione]

We're going to need this logic in a couple from shell tests too.

Should we:

1. Duplicate the logic into lit.cfg.py?
2. Import lldbplatformutil.py into lit.cfg.py? (are there any python structure/layout issues that would complicate this?)
3. Put the logic in an independent location and have both lit.cfg.py and lldbplatformutil.py import that?

[kastiglione]

To start with option #​1, I created #84246

[bulbazord]

To start with option #​1, I created #84246

Just took a look, thanks for taking care of that. I'll address your feedback (or just look at what you did in the other PR) and update this tomorrow.

[github-actions]

✅ With the latest revision this PR passed the Python code formatter.

[kastiglione]

looks good