Adapt TestCustomShell and TestMultipleDebuggers to run under ASAN

[augusto2112]

Summary

In situations where only LLDB is ASANified, a false positive occurs
unless ASAN_OPTIONS=detect_container_overflow=0 is set in the
environment.

[JDevlieghere]

Why are only these two tests affected? Should this be something we set globally for all the tests? The API tests already have support for forwarding ASAN_OPTIONS and lit has a similar concept.

I can imagine why TestMultipleSimultaneousDebuggers would be special: it basically build its own driver (so the detest machinery to set the inferior env doesn't work). But I'm a little puzzled by the shell tests, especially since you only updated a single RUN line.

[augusto2112]

In D143772#4119213, @JDevlieghere wrote:

Why are only these two tests affected? Should this be something we set globally for all the tests? The API tests already have support for forwarding ASAN_OPTIONS and lit has a similar concept.

Because they both set their own environment in one way or the other, which doesn't happen in the usual cas.

But I'm a little puzzled by the shell tests, especially since you only updated a single RUN line.

"env -i " means only use the enviroments passed in that line (discard everything else).

[aprantl]

Generally LGTM, with one possible improvement inline!

lldb/test/API/api/multiple-debuggers/TestMultipleDebuggers.py

20

IIUC, the python process executing this python code should already have this flag passed in via lit, so if it's possible to say something like 'ASAN_OPTIONS':getEnv('ASAN_OPTIONS' then that would be the cleanest implementation.

lldb/test/Shell/Host/TestCustomShell.test

10

This is as good as we can do.

[aprantl]

Why are only these two tests affected? Should this be something we set globally for all the tests? The API tests already have support for forwarding ASAN_OPTIONS and lit has a similar concept.

We do, but only for dotest tests, and our dotest test launches another LLDB.

[JDevlieghere]

In D143772#4119227, @augusto2112 wrote:

"env -i " means only use the enviroments passed in that line (discard everything else).

Thanks, I didn't notice -i.

LGTM

lldb/test/API/api/multiple-debuggers/TestMultipleDebuggers.py

 
 class TestMultipleSimultaneousDebuggers(TestBase):
 
     NO_DEBUG_INFO_TESTCASE = True
 
     @skipIfNoSBHeaders
     @skipIfWindows
     def test_multiple_debuggers(self):
-        env = {self.dylibPath: self.getLLDBLibraryEnvVal()}
+        env = {self.dylibPath: self.getLLDBLibraryEnvVal(),
+              # We need this in order to run under ASAN, in case only LLDB is ASANified.
+              'ASAN_OPTIONS': os.getenv('ASAN_OPTIONS', None)}
 
         self.driver_exe = self.getBuildArtifact("multi-process-driver")
         self.buildDriver('multi-process-driver.cpp', self.driver_exe)
         self.addTearDownHook(lambda: os.remove(self.driver_exe))
         self.signBinary(self.driver_exe)
 
         self.inferior_exe = self.getBuildArtifact("testprog")
         self.buildDriver('testprog.cpp', self.inferior_exe)

lldb/test/Shell/Host/TestCustomShell.test

 # FIXME: ShellExpandArguments is unimplemented on the following targets.
 # UNSUPPORTED: system-windows
 # UNSUPPORTED: system-linux
 # XFAIL: system-freebsd
 # XFAIL: system-netbsd
 # XFAIL: system-openbsd
 
 # RUN: %clang_host %S/Inputs/simple.c -g -o %t.out
 # RUN: SHELL=bogus not %lldb %t.out -b -o 'process launch -X 1 --' 2>&1 | FileCheck %s --check-prefix ERROR
-# RUN: env -i %lldb %t.out -b -o 'process launch -X 1 --' 2>&1 | FileCheck %s
+# RUN: env -i ASAN_OPTIONS='detect_container_overflow=0' %lldb %t.out -b -o 'process launch -X 1 --' 2>&1 | FileCheck %s
 
 # ERROR: error: shell expansion failed
 # CHECK-NOT: error: shell expansion failed