I've the sample pass code from LLVM.org:
#include "llvm/Pass.h"
#include "llvm/IR/Function.h"
#include "llvm/Support/raw_ostream.h"
using namespace llvm;
namespace {
struct Hello : public FunctionPass {
static char ID;
Hello() : FunctionPass(ID) {}
bool runOnFunction(Function &F) override {
errs() << "Hello: ";
errs().write_escaped(F.getName()) << '\n';
return false;
}
}; // end of struct Hello
} // end of anonymous namespace
char Hello::ID = 0;
static RegisterPass<Hello> X("hello", "Hello World Pass",
false /* Only looks at CFG */,
false /* Analysis Pass */);
The project builds fine and creates a SkeletonPass.dll.
When I execute the command:
C:\Users\nlykkei\Projects\llvm-pass-tutorial\build>opt -load skeleton\Debug\SkeletonPass.dll -hello foo.bc
opt: Unknown command line argument '-hello'. Try: 'opt -help'
opt: Did you mean '-help'?
opt doesn't recognize -hello option, even thus everything works fine on Ubuntu 16.04.
In addition, if I execute:
clang -Xclang -load -Xclang skeleton\Debug\SkeletonPass.dll foo.bc
nothing is printed out on Visual Studio terminal (Native Tools Command Prompt x86). On Linux, the function names are printed nicely for the same bitcode file.
What can be the reason for my experience? I do exactly the same on Windows 10 as I do on Ubuntu, but very different results.
Plugins are special beasts on Windows, because the latter does not support proper dynamic linking, so, your pass simply does not register itself in the PassRegistry. So you'd either need to compile all the LLVM into .dll or link your pass statically into opt / clang.
Related
I've always thought that using std::cout << something was thread safe.
For this little example
#include <iostream>
#include <thread>
void f()
{
std::cout << "Hello from f\n";
}
void g()
{
std::cout << "Hello from g\n";
}
int main()
{
std::thread t1(f);
std::thread t2(g);
t1.join();
t2.join();
}
my expectation was that the order of the two outputs would be undefined (and indeed that is what I observe in practice), but that the calls to operator<< are thread safe.
However, ThreadSanitizer, DRD and Helgrind all seem to give various errors regarding access to std::__1::ios_base::width(long) and std::__1::basic_ios<char, std::__1::char_traits >::fill()
On Compiler Explorer I don't see any errors.
On FreeBSD 13, ThreadSanitizer gives me 3 warnings, the two listed above plus the malloc/memcpy to the underlying i/o buffer.
Again in FreeBSD 13, DRD gives 4 errors, width() and fill() times two for the two threads.
Finally FreeBSD 13 Helgrind gives one known false positive related to TLS in thread creation, fill()and width() twice.
On Fedora 34
No errors with g++ 11.2.1 and ThreadSanitizer
DRD complains about malloc/memcpy in fwrite with g++ compiled exe
Helgrind also complains about fwrite and also for the construction of cout, again with the g++ compiled exe
clang++ 12 ThreadSanitizer complains about fill() and width()
DRD with the clang++ compiler exe complains about fill(), width(), fwrite and one other in start_thread
Helgrind with the clang++ exe complains about some TLS, fill(), width(), fwrite
macOS XCode clang++ ThreadSanitizer generates warnings as well (which will be libc++).
Looking at the libc++ and libstdc++ code I don't see anything at all that protects width(). So I don't understand why there are no complaints on compiler explorer.
I tried running with TSAN_OPTIONS=print_suppressions=1 and there was no more output (g++ Fedora ThreadSanitizer)
There does seem to be some consensus over the width() and fill() calls.
Looking more closely at the libstdc++ source I see that there is
(with some trimming and comments):
// ostream_insert.h
// __n is the length of the string pointed to by __s
template<typename _CharT, typename _Traits>
basic_ostream<_CharT, _Traits>&
__ostream_insert(basic_ostream<_CharT, _Traits>& __out,
const _CharT* __s, streamsize __n)
{
typedef basic_ostream<_CharT, _Traits> __ostream_type;
typedef typename __ostream_type::ios_base __ios_base;
typename __ostream_type::sentry __cerb(__out);
if (__cerb)
{
__try
{
const streamsize __w = __out.width();
if (__w > __n)
{
// snipped
// handle padding
}
else
__ostream_write(__out, __s, __n);
// why no hazard here?
__out.width(0);
}
__out is the stream object, global cout in this case. I don't see anything like locks or atomics.
Any suggestions as to how ThreadSanitizer/g++ is getting a "clean" output?
There is this somewhat cryptic comment
template<typename _CharT, typename _Traits>
basic_ostream<_CharT, _Traits>::sentry::
sentry(basic_ostream<_CharT, _Traits>& __os)
: _M_ok(false), _M_os(__os)
{
// XXX MT
if (__os.tie() && __os.good())
__os.tie()->flush();
The libc++ code looks similar. In iostream
template<class _CharT, class _Traits>
basic_ostream<_CharT, _Traits>&
__put_character_sequence(basic_ostream<_CharT, _Traits>& __os,
const _CharT* __str, size_t __len)
{
#ifndef _LIBCPP_NO_EXCEPTIONS
try
{
#endif // _LIBCPP_NO_EXCEPTIONS
typename basic_ostream<_CharT, _Traits>::sentry __s(__os);
if (__s)
{
typedef ostreambuf_iterator<_CharT, _Traits> _Ip;
if (__pad_and_output(_Ip(__os),
__str,
(__os.flags() & ios_base::adjustfield) == ios_base::left ?
__str + __len :
__str,
__str + __len,
__os,
__os.fill()).failed())
__os.setstate(ios_base::badbit | ios_base::failbit);
and in locale
template <class _CharT, class _OutputIterator>
_LIBCPP_HIDDEN
_OutputIterator
__pad_and_output(_OutputIterator __s,
const _CharT* __ob, const _CharT* __op, const _CharT* __oe,
ios_base& __iob, _CharT __fl)
{
streamsize __sz = __oe - __ob;
streamsize __ns = __iob.width();
if (__ns > __sz)
__ns -= __sz;
else
__ns = 0;
for (;__ob < __op; ++__ob, ++__s)
*__s = *__ob;
for (; __ns; --__ns, ++__s)
*__s = __fl;
for (; __ob < __oe; ++__ob, ++__s)
*__s = *__ob;
__iob.width(0);
return __s;
}
Again I see no thread protection, but also this time the tools detect a hazard.
Are these real issues? For plain calls to operator<< the value of width doesn't change, and is always 0.
libstdc++ does not produce the error while libc++ does.
iostream.objects.overview Concurrent access to a synchronized ([ios.members.static]) standard iostream object's formatted and unformatted input ([istream]) and output ([ostream]) functions or a standard C stream by multiple threads does not result in a data race ([intro.multithread]).
So this looks like a libc++ bug to me.
I got the answer from Jonathan Wakely. Makes me feel rather stupid.
The difference is that on Fedora, libstdc++.so contains an explicit instantiation of the iostream classes. libstdc++.so isn't instrumented for ThreadSanitizer so it cannot detect any hazards related to it.
Is it possible to side-step _NSGetExecutablePath on Ubuntu Linux in place of a non-Apple specific approach?
I am trying to compile the following code on Ubuntu: https://github.com/Bohdan-Khomtchouk/HeatmapGenerator/blob/master/HeatmapGenerator2_Macintosh_OSX.cxx
As per this prior question that I asked: fatal error: mach-o/dyld.h: No such file or directory, I decided to comment out line 52 and am wondering if there is a general cross-platform (non-Apple specific) way that I can rewrite the code block of line 567 (the _NSGetExecutablePath block) in a manner that is non-Apple specific.
Alen Stojanov's answer to Programmatically retrieving the absolute path of an OS X command-line app and also How do you determine the full path of the currently running executable in go? gave me some ideas on where to start but I want to make certain that I am on the right track here before I go about doing this.
Is there a way to modify _NSGetExecutablePath to be compatible with Ubuntu Linux?
Currently, I am experiencing the following compiler error:
HeatmapGenerator_Macintosh_OSX.cxx:568:13: error: use of undeclared identifier
'_NSGetExecutablePath'
if (_NSGetExecutablePath(path, &size) == 0)
Basic idea how to do it in a way that should be portable across POSIX systems:
#define _XOPEN_SOURCE 500
#include <stdio.h>
#include <limits.h>
#include <stdlib.h>
static char *path;
const char *appPath(void)
{
return path;
}
static void cleanup()
{
free(path);
}
int main(int argc, char **argv)
{
path = realpath(argv[0], 0);
if (!path)
{
perror("realpath");
return 1;
}
atexit(&cleanup);
printf("App path: %s\n", appPath());
return 0;
}
You can define an own module for it, just pass it argv[0] and export the appPath() function from a header.
edit: replaced exported variable by accessor method
OK, I admit it, this is a unique case. When we build our application we are using make, so I've included my tests in a test folder under src. Then at the same level as our release folder we have created a unit-test folder that includes all of our source files and our test source files.
But my IDE is CLion, which uses CMake. In my CMakeLists.txt file, I have included:
enable_testing()
find_package(GTest REQUIRED)
include_directories(${GTEST_INCLUDE_DIRS})
add_executable(TestProject ${SOURCE_FILES})
target_link_libraries(TestProject ${GTEST_BOTH_LIBRARIES})
I am creating my first Test Fixture. Here is the code:
#include "OPProperties.h"
#include "gtest/gtest.h"
namespace {
// The fixture for testing class OPPropertiesTestTest.
class OPPropertiesTestTest : public ::testing::Test {
protected:
// You can remove any or all of the following functions if its body
// is empty.
OPPropertiesTestTest() {
// You can do set-up work for each test here.
}
virtual ~OPPropertiesTestTest() {
// You can do clean-up work that doesn't throw exceptions here.
}
// If the constructor and destructor are not enough for setting up
// and cleaning up each test, you can define the following methods:
virtual void SetUp() {
// Code here will be called immediately after the constructor (right
// before each test).
}
virtual void TearDown() {
// Code here will be called immediately after each test (right
// before the destructor).
}
// Objects declared here can be used by all tests in the test case for OPPropertiesTestTest.
};
TEST_F(OPPropertiesTestTest, ThisTestWillFail) {
EXPECT_EQ(0, 2);
}
} // namespace
int main(int argc, char **argv) {
::testing::InitGoogleTest(&argc, argv);
return RUN_ALL_TESTS();
}
Here is an image capture:
Notice the syntax checker errors in my TEST_F function. When I started to type TEST_F code completion is trying to find a Boost Test Function.
Can someone tell me what else I need to add to the CMakeLists.txt file or what I am not doing that GTest functions are not being recognized?
As πάντα ῥεῖ pointed out, I hadn't actually tried to build the code. When I did I first received a linker error for pthread, so we added the following line the the CMakeLists.txt file:
target_link_libraries(OnePrint pthread)
Then I tried again to build and received these errors:
/home/user/gtest-1.7.0/lib/.libs/libgtest.so: undefined reference to `pthread_key_create'
/home/user/gtest-1.7.0/lib/.libs/libgtest.so: undefined reference to `pthread_getspecific'
/home/user/gtest-1.7.0/lib/.libs/libgtest.so: undefined reference to `pthread_key_delete'
/home/user/gtest-1.7.0/lib/.libs/libgtest.so: undefined reference to `pthread_setspecific'
collect2: error: ld returned 1 exit status
So, I ran a search on these errors and found this question.
The answer that worked for me was here.
I am trying to make our program runnable on some old Linux versions. One common import that prevents it is __longjmp_chk, added in glibc 2.11 but missing in older ones. One "solution" is to use -D_FORTIFY_SOURCE=0 but this turns off other fortify functions (__printf_chk etc) which are present in the target libc. Is there a way to make __longjmp_chk a "weak import" which would use the function from libc.so.6 if present, and fall back to local stub if not?
Is there a way to make __longjmp_chk a "weak import" which would use
the function from libc.so.6 if present, and fall back to local stub
if not?
I'd say yes, using dlsym() to check for __longjmp_chk and acting accordingly:
/* cc -ldl */
#define _GNU_SOURCE
#include <setjmp.h>
#include <stdio.h>
#include <dlfcn.h>
void __longjmp_chk(sigjmp_buf env, int val)
{
void (*p)(sigjmp_buf, int) = dlsym(RTLD_NEXT, "__longjmp_chk");
if (p)
printf("use the function from libc\n"),
p(env, val);
else
{
printf("falling back to local stub\n");
/* local stub - whatever that may be */
}
}
main()
{ // try it
sigjmp_buf env;
while (!setjmp(env)) __longjmp_chk(env, 1);
return 0;
}
I am trying to make our program runnable on some old Linux versions.
There are only a few ways to make this work, and most of them are enumerated here.
Is there a way to make __longjmp_chk a "weak import".
No.
I have a shared library "libwiston.so". I am using this to create another shared library called "libAnimation.so", which will be used by another project. Now, the second library "libAnimation.so" can't be used in test code correctly. So I doubt that the creation of the second lib "libAnimation.so" is right. The gcc command to create this lib is
g++ -g -shared -Wl,-soname,libwiston.so -o libAnimation.so $(objs) -lc".
Has someone come across this problem?
That looks like a weird link line - you are creating libAnimation.so, but its internal DT_SONAME name is libwiston.so.
I don't think that what you wanted to do. Don't you want to link libAnimation.so against libwiston.so (-lwiston)?
g++ -g -shared -o libAnimation.so $(objs) -lc -lwiston
I think it would be easier to wrap your build in automake/autoconf and rely on libtool to get the shared library creation correct.
I'll do a humble review on the process of creating shared libraries.
Let's begin by creating libwiston.so. First we implement the function we would like to export and then define it on a header so other programs knows how to call it.
/* file libwiston.cpp
* Implementation of hello_wiston(), called by libAnimation.so
*/
#include "libwiston.h"
#include <iostream>
int hello_wiston(std::string& msg)
{
std::cout << msg << std::endl;
return 0;
}
and:
/* file libwiston.h
* Exports hello_wiston() as a C symbol.
*/
#include <string>
extern "C" {
int hello_wiston(std::string& msg);
};
This code can be compiled with: g++ libwiston.cpp -o libwiston.so -shared
Now we implement the second shared library, named libAnimation.so that calls the function exported by the first library.
/* file libAnimation.cpp
* Implementation of call_wiston().
* This function is a simple wrapper around hello_wiston().
*/
#include "libAnimation.h"
#include "libwiston.h"
#include <iostream>
int call_wiston(std::string& param)
{
hello_wiston(param);
return 0;
}
and header:
/* file libAnimation.h
* Exports call_wiston() as a C symbol.
*/
#include <string>
extern "C" {
int call_wiston(std::string& param);
};
Compile it with: g++ libAnimation.cpp -o libAnimation.so -shared -L. -lwiston
Finally, we create a small application to test libAnimation.
/* file demo.cpp
* Implementation of the test application.
*/
#include "libAnimation.h"
int main()
{
std::string msg = "hello stackoverflow!";
call_wiston(msg);
}
And compile it with: g++ demo.cpp -o demo -L. -lAnimation
There's an interesting tool named nm that you can use to list the symbols exported by your shared library. Using these examples, you could execute the following commands to check for the symbols:
nm libAnimation.so | grep call_wiston
outputs:
00000634 t _GLOBAL__I_call_wiston
000005dc T call_wiston
and also:
nm libwiston.so | grep hello_wiston
outputs:
0000076c t _GLOBAL__I_hello_wiston
000006fc T hello_wiston