[lldb] convert jit-loader_rtdyld_elf.test to an API test #170333
Conversation
|
@llvm/pr-subscribers-platform-windows @llvm/pr-subscribers-lldb Author: Charles Zablit (charles-zablit) ChangesThis patch converts the This test is timing out in CI on Windows and the hang cannot be reproduced at desk. Converting it to an API test would allow us to instrument it better in order to trace the failure. Full diff: https://github.com/llvm/llvm-project/pull/170333.diff 4 Files Affected:
diff --git a/lldb/test/API/functionalities/breakpoint/jit_loader_rtdyld_elf/Makefile b/lldb/test/API/functionalities/breakpoint/jit_loader_rtdyld_elf/Makefile
new file mode 100644
index 0000000000000..51a6bfc5fee67
--- /dev/null
+++ b/lldb/test/API/functionalities/breakpoint/jit_loader_rtdyld_elf/Makefile
@@ -0,0 +1,14 @@
+CXX_SOURCES := jitbp.cpp
+
+include Makefile.rules
+
+# Build LLVM IR that lli will JIT.
+# Matches the commands from the shell test.
+jitbp.ll: jitbp.cpp
+ $(CXX) -g -S -emit-llvm --target=x86_64-unknown-unknown-elf \
+ -o $@ $<
+
+all: jitbp.ll
+
+clean::
+ rm -f jitbp.ll
diff --git a/lldb/test/API/functionalities/breakpoint/jit_loader_rtdyld_elf/TestJitBreakPoint.py b/lldb/test/API/functionalities/breakpoint/jit_loader_rtdyld_elf/TestJitBreakPoint.py
new file mode 100644
index 0000000000000..1a4b16f03f474
--- /dev/null
+++ b/lldb/test/API/functionalities/breakpoint/jit_loader_rtdyld_elf/TestJitBreakPoint.py
@@ -0,0 +1,53 @@
+"""
+Test that pending breakpoints resolve for JITted code with mcjit and rtdyld.
+"""
+
+import lldb
+from lldbsuite.test.decorators import *
+from lldbsuite.test.lldbtest import *
+from lldbsuite.test import lldbutil
+
+
+class TestJitBreakpoint(TestBase):
+ def setUp(self):
+ TestBase.setUp(self)
+ self.ll = self.getBuildArtifact("jitbp.ll")
+
+ @skipUnlessArch("x86_64")
+ @expectedFailureAll(oslist=["windows"])
+ def test_jit_breakpoints(self):
+ self.build()
+ self.do_test("--jit-kind=mcjit")
+ self.do_test("--jit-linker=rtdyld")
+
+ def do_test(self, jit_flag: str):
+ self.dbg.SetAsync(False)
+
+ self.dbg.HandleCommand("settings set plugin.jit-loader.gdb.enable on")
+
+ lli_path = lldbutil.which("lli")
+ self.assertIsNotNone(lli_path, "Could not find lli executable")
+ target = self.dbg.CreateTarget(lli_path)
+ self.assertTrue(target.IsValid())
+
+ bp = target.BreakpointCreateByName("jitbp")
+ self.assertTrue(bp.IsValid())
+ self.assertEqual(bp.GetNumLocations(), 0, "Expected a pending breakpoint")
+
+ launch_info = target.GetLaunchInfo()
+ launch_info.SetArguments([jit_flag, self.ll], True)
+
+ error = lldb.SBError()
+ process = target.Launch(launch_info, error)
+ self.assertTrue(process.IsValid())
+ self.assertTrue(error.Success(), error.GetCString())
+
+ self.assertEqual(process.GetState(), lldb.eStateStopped)
+
+ thread = process.GetSelectedThread()
+ frame = thread.GetSelectedFrame()
+ self.assertIn("jitbp", frame.GetFunctionName())
+
+ self.assertGreaterEqual(
+ bp.GetNumLocations(), 1, "Breakpoint must be resolved after JIT loads code"
+ )
diff --git a/lldb/test/Shell/Breakpoint/Inputs/jitbp.cpp b/lldb/test/API/functionalities/breakpoint/jit_loader_rtdyld_elf/jitbp.cpp
similarity index 100%
rename from lldb/test/Shell/Breakpoint/Inputs/jitbp.cpp
rename to lldb/test/API/functionalities/breakpoint/jit_loader_rtdyld_elf/jitbp.cpp
diff --git a/lldb/test/Shell/Breakpoint/jit-loader_rtdyld_elf.test b/lldb/test/Shell/Breakpoint/jit-loader_rtdyld_elf.test
deleted file mode 100644
index ae9402a519494..0000000000000
--- a/lldb/test/Shell/Breakpoint/jit-loader_rtdyld_elf.test
+++ /dev/null
@@ -1,22 +0,0 @@
-# REQUIRES: target-x86_64
-# XFAIL: system-windows
-
-# RuntimeDyld can be used to link and load emitted code for both, MCJIT and Orc.
-#
-# RUN: %clangxx -g -S -emit-llvm --target=x86_64-unknown-unknown-elf \
-# RUN: -o %t.ll %p/Inputs/jitbp.cpp
-#
-# RUN: %lldb -b -o 'settings set plugin.jit-loader.gdb.enable on' -o 'b jitbp' \
-# RUN: -o 'run --jit-kind=mcjit %t.ll' lli | FileCheck %s
-#
-# RUN: %lldb -b -o 'settings set plugin.jit-loader.gdb.enable on' -o 'b jitbp' \
-# RUN: -o 'run --jit-linker=rtdyld %t.ll' lli | FileCheck %s
-
-# CHECK: Breakpoint 1: no locations (pending).
-# CHECK: (lldb) run {{.*}}
-# CHECK: Process {{.*}} launched: {{.*}}
-# CHECK: Process {{.*}} stopped
-# CHECK: JIT(0x{{.*}})`jitbp() at jitbp.cpp:1:15
-# CHECK: -> 1 int jitbp() { return 0; }
-# CHECK: ^
-# CHECK: 2 int main() { return jitbp(); }
|
🐧 Linux x64 Test Results
✅ The build succeeded and all tests passed. |
charles-zablit
force-pushed
the
charles-zablit/lldb/convert-shell-test-to-api-test
branch
from
77c1872 to
29ee61c
Compare
|
✅ With the latest revision this PR passed the Python code formatter. |
charles-zablit
force-pushed
the
charles-zablit/lldb/convert-shell-test-to-api-test
branch
2 times, most recently
from
c83a241 to
344dfcc
Compare
charles-zablit
force-pushed
the
charles-zablit/lldb/convert-shell-test-to-api-test
branch
from
344dfcc to
07ecfc8
Compare
| TestBase.setUp(self) | ||
| self.ll = self.getBuildArtifact("jitbp.ll") | ||
|
|
||
| @skipUnlessArch("x86_64") |
There was a problem hiding this comment.
We'll also need a decorator that checks that the compiler is clang (and not, e.g., gcc) since we use the -emit-llvm flag.
There was a problem hiding this comment.
I have added @skipUnlessCompilerIsClang
| class TestJitBreakpoint(TestBase): | ||
| def setUp(self): | ||
| TestBase.setUp(self) | ||
| self.ll = self.getBuildArtifact("jitbp.ll") |
There was a problem hiding this comment.
If you make this a local variable below, you can remove the entire setUp function.
There was a problem hiding this comment.
Fixed, thanks 👍
| from lldbsuite.test.decorators import * | ||
| from lldbsuite.test.lldbtest import * | ||
|
|
||
| import shutil |
There was a problem hiding this comment.
Maybe sinking this into the function below makes test discovery marginally faster?
There was a problem hiding this comment.
Fixed, thanks 👍
| error = lldb.SBError() | ||
| process = target.Launch(launch_info, error) | ||
| self.assertTrue(process.IsValid()) | ||
| self.assertTrue(error.Success(), error.GetCString()) |
There was a problem hiding this comment.
is there a good reason why we can't use any of the lldbutil helper functions that launch a program and check for breakpoints here?
There was a problem hiding this comment.
At first, I tried to mimic the Shell test style. Now, I copied how the TestInlinedBreakpoints.py test is setup.
|
|
||
| lldb_dir = os.path.dirname(lldbtest_config.lldbExec) | ||
| lli_path = shutil.which("lli", path=lldb_dir) | ||
| self.assertTrue(os.path.exists(lli_path), "lli not found") |
There was a problem hiding this comment.
This is going to fail in out-of-tree builds?
Do we have a way to get the LLVM BINDIR here?
There was a problem hiding this comment.
I have replaced the whole invocation with self.runCmd("target create lli", CURRENT_EXECUTABLE_SET). This is what we were doing in the shell version of the test.
I did not find a way to get the LLVM bindir.
| self.assertTrue(clang_path, "built clang could not be found") | ||
| lli_path = os.path.join(os.path.dirname(clang_path), "lli") | ||
| self.assertTrue(lldbutil.is_exe(lli_path), f"'{lli_path}' is not an executable") | ||
| self.runCmd(f"target create {lli_path}", CURRENT_EXECUTABLE_SET) |
There was a problem hiding this comment.
lldbutil.run_to_source_breakpoint takes the path to the executable to run as a named argument, and a boolean for whether the pattern is found on launch, so you could replace from line 27 to line 46 with:
lldbutil.run_to_source_breakpoint(self, "int jitbp()", lldb.SBFileSpec("jitbt.cpp"), exe_name=lli_path, has_locations_before_run=False)
There was a problem hiding this comment.
But if your goal is to have this be in separate pieces in case you need to instrument it when it mysteriously fails on some bot, maybe doing the separate stages by hand in the test is okay too.
There was a problem hiding this comment.
I think that, as you said, we should keep the separate stages to instrument the test later on once it times out.
4 participants
This patch converts the
jit-loader_rtdyld_elf.testtest from a Shell test to an API test.This test is timing out in CI on Windows and the hang cannot be reproduced at desk. Converting it to an API test would allow us to instrument it better in order to trace the failure.