[compiler-rt] r355236 - [sanitizers] Don't use Windows Trace Logging on MinGW
Volodymyr Sapsai via llvm-commits
llvm-commits at lists.llvm.org
Fri Mar 1 18:18:11 PST 2019
Thanks for checking this and for the revert.
> On Mar 1, 2019, at 17:33, Evgenii Stepanov <eugenis at google.com> wrote:
>
> r355231 reverted.
>
> On Fri, Mar 1, 2019 at 5:28 PM Evgenii Stepanov <eugenis at google.com> wrote:
>>
>> Well, realistically, nothing we do at this point can make this code any uglier.
>> Might as well add an argument to COMMON_INTERCEPTOR_ENTER that would
>> evaluate to "return" for interceptors with a non-void return value,
>> and to "" otherwise :(
>>
>> I'll revert for now.
>>
>> On Fri, Mar 1, 2019 at 5:11 PM Volodymyr Sapsai <vsapsai at apple.com> wrote:
>>>
>>> And the error is
>>>
>>> In file included from /Users/buildslave/jenkins/workspace/clang-stage1-configure-RA/llvm/projects/compiler-rt/lib/tsan/rtl/tsan_interceptors.cc:2351:
>>>
>>> /Users/buildslave/jenkins/workspace/clang-stage1-configure-RA/llvm/projects/compiler-rt/lib/tsan/../sanitizer_common/sanitizer_common_interceptors.inc:5538:3: error: void function 'wrap___bzero' should not return a value [-Wreturn-type]
>>>
>>> COMMON_INTERCEPTOR_MEMSET_IMPL(ctx, block, 0, size);
>>> ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>> /Users/buildslave/jenkins/workspace/clang-stage1-configure-RA/llvm/projects/compiler-rt/lib/tsan/../sanitizer_common/sanitizer_common_interceptors.inc:246:5: note: expanded from macro 'COMMON_INTERCEPTOR_MEMSET_IMPL'
>>> COMMON_INTERCEPTOR_ENTER(ctx, memset, dst, v, size); \
>>> ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>> /Users/buildslave/jenkins/workspace/clang-stage1-configure-RA/llvm/projects/compiler-rt/lib/tsan/rtl/tsan_interceptors.cc:2241:3: note: expanded from macro 'COMMON_INTERCEPTOR_ENTER'
>>> SCOPED_TSAN_INTERCEPTOR(func, __VA_ARGS__); \
>>> ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>>> /Users/buildslave/jenkins/workspace/clang-stage1-configure-RA/llvm/projects/compiler-rt/lib/tsan/rtl/tsan_interceptors.h:49:7: note: expanded from macro 'SCOPED_TSAN_INTERCEPTOR'
>>> return REAL(func)(__VA_ARGS__); \
>>> ^ ~~~~~~~~~~~~~~~~~~~~~~~
>>>
>>>
>>>
>>> On Mar 1, 2019, at 17:07, Volodymyr Sapsai via llvm-commits <llvm-commits at lists.llvm.org> wrote:
>>>
>>> There are 2 failing bots:
>>> http://green.lab.llvm.org/green/job/clang-stage1-configure-RA/
>>> http://green.lab.llvm.org/green/job/clang-stage1-cmake-RA-expensive/
>>> The first one is more useful as the second one is much slower.
>>>
>>> Compiler diagnostic seems to be correct but unfortunately I don’t see a simple way to fix tsan interceptors.
>>>
>>> On Mar 1, 2019, at 17:02, Vlad Tsyrklevich <vlad at tsyrklevich.net> wrote:
>>>
>>> Volodymyr, your change fixed the build on the Linux sanitizer bots. I originally misread the buildbot data and thought it hadn't. I have no idea about macOS. I've pinged Evgenii to take a look at this thread, is there a broken bot you can point him to?
>>>
>>> On Fri, Mar 1, 2019 at 4:46 PM Volodymyr Sapsai <vsapsai at apple.com> wrote:
>>>>
>>>> Does your revert help with the build? Because we are still seeing compilation errors on macOS, presumably caused by r355231 https://github.com/llvm/llvm-project/commit/ddc4b7c1d6d4d4ffd8538b40c00c3fb16b1b1ab0
>>>>
>>>>
>>>> On Mar 1, 2019, at 16:41, Vlad Tsyrklevich <vlad at tsyrklevich.net> wrote:
>>>>
>>>> I've reverted both of these changes in r355250. The sanitizer build fails on Linux and I have other failures to investigate. (For what it's worth Volodymyr, I took a brief look and it was unclear to me if either were correct. The Linux build was failing before and after your change.)
>>>>
>>>> On Fri, Mar 1, 2019 at 4:07 PM Volodymyr Sapsai via llvm-commits <llvm-commits at lists.llvm.org> wrote:
>>>>>
>>>>> Martin, I’ve made a small fix in r355244. See inline for more details
>>>>>
>>>>>> On Mar 1, 2019, at 14:30, Martin Storsjo via llvm-commits <llvm-commits at lists.llvm.org> wrote:
>>>>>>
>>>>>> Author: mstorsjo
>>>>>> Date: Fri Mar 1 14:30:14 2019
>>>>>> New Revision: 355236
>>>>>>
>>>>>> URL: http://llvm.org/viewvc/llvm-project?rev=355236&view=rev
>>>>>> Log:
>>>>>> [sanitizers] Don't use Windows Trace Logging on MinGW
>>>>>>
>>>>>> mingw-w64 currently is lacking the headers for this feature.
>>>>>>
>>>>>> Make the include lowercase at the same time. We consistently
>>>>>> use lowercase for windows header includes, as windows itself is
>>>>>> case insensitive, the SDK headers (in general, not necessarily
>>>>>> considering this particular header) aren't consistent among themselves
>>>>>> about what the proper canonical capitalization for headers are,
>>>>>> and MinGW uses all lowercase names for the headers (as it is often
>>>>>> used on case sensitive filesystems).
>>>>>>
>>>>>> In case mingw-w64 later gets this header, we can revert this
>>>>>> (but keep the include lowercased).
>>>>>>
>>>>>> Differential Revision: https://reviews.llvm.org/D58765
>>>>>>
>>>>>> Modified:
>>>>>> compiler-rt/trunk/lib/sanitizer_common/sanitizer_common.h
>>>>>> compiler-rt/trunk/lib/sanitizer_common/sanitizer_win.cc
>>>>>>
>>>>>> Modified: compiler-rt/trunk/lib/sanitizer_common/sanitizer_common.h
>>>>>> URL: http://llvm.org/viewvc/llvm-project/compiler-rt/trunk/lib/sanitizer_common/sanitizer_common.h?rev=355236&r1=355235&r2=355236&view=diff
>>>>>> ==============================================================================
>>>>>> --- compiler-rt/trunk/lib/sanitizer_common/sanitizer_common.h (original)
>>>>>> +++ compiler-rt/trunk/lib/sanitizer_common/sanitizer_common.h Fri Mar 1 14:30:14 2019
>>>>>> @@ -804,7 +804,13 @@ enum AndroidApiLevel {
>>>>>>
>>>>>> void WriteToSyslog(const char *buffer);
>>>>>>
>>>>>> -#if SANITIZER_MAC || SANITIZER_WINDOWS
>>>>>> +#if defined(SANITIZER_WINDOWS) && defined(_MSC_VER)
>>>>>> +#define SANITIZER_WIN_TRACE 1
>>>>>> +#else
>>>>>> +#define SANITIZER_WIN_TRACE 0
>>>>>> +#endif
>>>>>> +
>>>>>> +#if SANITIZER_LINUX || SANITIZER_WIN_TRACE
>>>>> I think changing SANITIZER_MAC to SANITIZER_LINUX was a mistake and was copy-pasted from the part of the file below. But I can be wrong and maybe we need #if SANITIZER_MAC || SANITIZER_LINUX || SANITIZER_WIN_TRACE
>>>>>
>>>>>> void LogFullErrorReport(const char *buffer);
>>>>>> #else
>>>>>> INLINE void LogFullErrorReport(const char *buffer) {}
>>>>>> @@ -818,7 +824,7 @@ INLINE void WriteOneLineToSyslog(const c
>>>>>> INLINE void LogMessageOnPrintf(const char *str) {}
>>>>>> #endif
>>>>>>
>>>>>> -#if SANITIZER_LINUX || SANITIZER_WINDOWS
>>>>>> +#if SANITIZER_LINUX || SANITIZER_WIN_TRACE
>>>>>> // Initialize Android logging. Any writes before this are silently lost.
>>>>>> void AndroidLogInit();
>>>>>> void SetAbortMessage(const char *);
>>>>>>
>>>>>> Modified: compiler-rt/trunk/lib/sanitizer_common/sanitizer_win.cc
>>>>>> URL: http://llvm.org/viewvc/llvm-project/compiler-rt/trunk/lib/sanitizer_common/sanitizer_win.cc?rev=355236&r1=355235&r2=355236&view=diff
>>>>>> ==============================================================================
>>>>>> --- compiler-rt/trunk/lib/sanitizer_common/sanitizer_win.cc (original)
>>>>>> +++ compiler-rt/trunk/lib/sanitizer_common/sanitizer_win.cc Fri Mar 1 14:30:14 2019
>>>>>> @@ -20,7 +20,6 @@
>>>>>> #include <io.h>
>>>>>> #include <psapi.h>
>>>>>> #include <stdlib.h>
>>>>>> -#include <TraceLoggingProvider.h>
>>>>>>
>>>>>> #include "sanitizer_common.h"
>>>>>> #include "sanitizer_file.h"
>>>>>> @@ -32,6 +31,8 @@
>>>>>> #if defined(PSAPI_VERSION) && PSAPI_VERSION == 1
>>>>>> #pragma comment(lib, "psapi")
>>>>>> #endif
>>>>>> +#if SANITIZER_WIN_TRACE
>>>>>> +#include <traceloggingprovider.h>
>>>>>> // Windows trace logging provider init
>>>>>> #pragma comment(lib, "advapi32.lib")
>>>>>> TRACELOGGING_DECLARE_PROVIDER(g_asan_provider);
>>>>>> @@ -39,6 +40,9 @@ TRACELOGGING_DECLARE_PROVIDER(g_asan_pro
>>>>>> TRACELOGGING_DEFINE_PROVIDER(g_asan_provider, "AddressSanitizerLoggingProvider",
>>>>>> (0x6c6c766d, 0x3846, 0x4e6a, 0xa4, 0xfb, 0x5b,
>>>>>> 0x53, 0x0b, 0xd0, 0xf3, 0xfa));
>>>>>> +#else
>>>>>> +#define TraceLoggingUnregister(x)
>>>>>> +#endif
>>>>>>
>>>>>> // A macro to tell the compiler that this part of the code cannot be reached,
>>>>>> // if the compiler supports this feature. Since we're using this in
>>>>>> @@ -1080,6 +1084,7 @@ u32 GetNumberOfCPUs() {
>>>>>> return sysinfo.dwNumberOfProcessors;
>>>>>> }
>>>>>>
>>>>>> +#if SANITIZER_WIN_TRACE
>>>>>> // TODO(mcgov): Rename this project-wide to PlatformLogInit
>>>>>> void AndroidLogInit(void) {
>>>>>> HRESULT hr = TraceLoggingRegister(g_asan_provider);
>>>>>> @@ -1103,6 +1108,7 @@ void LogFullErrorReport(const char *buff
>>>>>> TraceLoggingValue(buffer, "AsanReportContents"));
>>>>>> }
>>>>>> }
>>>>>> +#endif // SANITIZER_WIN_TRACE
>>>>>>
>>>>>> } // namespace __sanitizer
>>>>>>
>>>>>>
>>>>>>
>>>>>> _______________________________________________
>>>>>> llvm-commits mailing list
>>>>>> llvm-commits at lists.llvm.org
>>>>>> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>>>>
>>>>> _______________________________________________
>>>>> llvm-commits mailing list
>>>>> llvm-commits at lists.llvm.org
>>>>> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>>>
>>>>
>>>
>>> _______________________________________________
>>> llvm-commits mailing list
>>> llvm-commits at lists.llvm.org
>>> https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-commits
>>>
>>>
More information about the llvm-commits
mailing list