Please fix the hang in AsanTSDInit

Not bad. It looks like FreeBSD ASan broke too... https://bugs.llvm.org/show_bug.cgi?id=37238

In D46006#1077941, @ro wrote:

In D46006#1076822, @krytarowski wrote:

Can you run tests on SunOS? check-asan, check-ubsan etc.

Of course: that's the first thing I did after fixing the build (run make check-all, that is, which includes the above). Same as before the
breakage.

Can you run tests on SunOS? check-asan, check-ubsan etc.

Does it work for you for 32-bit? @joerg opposed adoption of 32-bit version as it seemed to be unimplemented..

This patch looks fine, not sure if we can improve it.

FreeBSD also uses pthread_np.h header.

Just a note that FreeBSD calls its function pthread_set_name_np.

This does not fix the potential problem on NetBSD. We define __VA_LIST_DECLARED (with extra _ on the beginning).

I cannot test it shortly on NetBSD, but if there are any issues left (probably none) - I will fix it in future.

In D43751#1031012, @emaste wrote:

In fact setting -z origin here on (current) versions of FreeBSD serves no purpose, but it also doesn't hurt anything.

ld(1) might accept -z origin, but why do we want it for other OSes? I don't see it used in the NetBSD dynamic ELF loader. DF_ORIGIN is just defined but not used by anything.

While there. There is feature in Clang/LLVM to print stacktrace on a crash. This is unusable on BSDs, as it prints signal handler's stacktrace instead of the program one. So regardless of cargo culting backtrace(3) it does not work as intended.

I've entirely switched to NetBSD kernel fixes in order to host fully-functional LLDB. Once I will be done with the debugger, I will be back to sanitizers and work on connecting a buildbot and remaining fixes and cleanup.

I've entirely switched to NetBSD kernel fixes in order to host fully-functional LLDB. Once I will be done with the debugger, I will be back to sanitizers and work on connecting a buildbot and remaining fixes and cleanup.

In D43751#1027234, @emaste wrote:

LGTM from the FreeBSD perspective

(Aside, it seems to me -Wl,-z,origin could be applied unconditionally, not only on FreeBSD/DragonFly).

Another failure example:

Knowing OpenBSD, their goal is no compatibility with others.

Please don't submit untested patches, they should be at least build-tested.

Reverted, broken syntax.

Please improve the description. List supported interceptors and changes in existing interceptors.

The title and description does not match the diff.

Hmm, this already landed as a part of D43080

::pthread_cancel is available for NetBSD as well.

Please try to partition this into 200-300 line patches.

Assuming that OpenBSD is capable to run only a subset of functionality of UBSan (no unwinder, no 32-bit on 64-bit kernel etc), do we need all the code like sanitizer_platform_limits_openbsd?

I would add an inline comment what's the purpose of this operation in CMake. Otherwise it looks good to me!

This looks like broke ASan on NetBSD:

I will add test name to CHECK in the test, and commit all three tests.

@vitalybuka I think you have reviewed an older snapshot. All the mentioned now issues are gone.

• apply patches from review

• improve the patch

I think that this should be a release blocker for 6.0.0. CC: @hans

• apply the remaining comments from review

• drop interceptors for setttyent(3) and endttyent(3)

• apply part of the comments from review

• apply comments from review
• improve the test

• add missing changes to sanitizer_platform_limits_netbsd
• fix a typo

apply comments from review

Ping?

This looks good to me!

In D42702#1018130, @bdrewery wrote:

This fixes it for me in the spirit of other library lookups (in that it doesn't bother with -L, but that could easily be added in). Despite Backtrace_LIBRARIES being plural I don't think it ever will be looking over FindBacktrace.cmake.

I have no clue about static linkage here.

https://people.freebsd.org/~bdrewery/patches/cmake-libexecinfo.diff

diff --git lib/Support/CMakeLists.txt lib/Support/CMakeLists.txt
index 5723f8fcf5b..dfbdd70701f 100644
--- lib/Support/CMakeLists.txt
+++ lib/Support/CMakeLists.txt
@@ -13,7 +13,9 @@ elseif( CMAKE_HOST_UNIX )
     set(system_libs ${system_libs} ${CMAKE_DL_LIBS})
   endif()
   if( HAVE_BACKTRACE )
-    set(system_libs ${system_libs} ${Backtrace_LIBRARIES})
+    get_filename_component(Backtrace_LIBFILE ${Backtrace_LIBRARIES} NAME_WE)
+    STRING(REGEX REPLACE "^lib" "" Backtrace_LIBFILE ${Backtrace_LIBFILE})
+    set(system_libs ${system_libs} ${Backtrace_LIBFILE})
   endif()
   if(LLVM_ENABLE_TERMINFO)
     if(HAVE_TERMINFO)

Hmm on the other hand.. these interceptors will be useful for NetBSD forks/descendants like FreeBSD (OpenBSD?) and perhaps other BSD descendants like SunOS (Solaris). I just don't enable them without prior-testing.

I can move the patches there. How about D43543, D43541, D43539? Can I commit as it is and move them in one patch to a new file.

ping?

This is considered to be a local patch.

Here is an approach with save_errno, useful in libraries when we preserve errno for public functions in the API of a library as-is, and not leaking it from the internals to unrelated code.

This just helps debugging connectivity issues in other unrelated code-parts (I'm debugging why select(2) / recv(2) don't receive a packet of type qHostInfo).

As discussed with Joerg, _PTHREADS is considered legacy and we can bear with lack of it and inconsistency with GCC.

Naively looking at https://hardenedbsd.org/content/easy-feature-comparison PIE on FreeBSD might need more work.

In D43168#1012853, @krytarowski wrote:

If I read the CMake's source code correctly, just including find_package(Threads) is enough to set OS specific flags.

https://github.com/Kitware/CMake/blob/9a509099f77ed32a0845e4e3fad7b8f1eb9be10b/Modules/FindThreads.cmake

If I read the CMake's source code correctly, just including find_package(Threads) is enough to set OS specific flags.

-pthreads defines additional OS specific flags like _REENTRANT or _PTHREADS and this is the form that should be used.

This also covers FreeBSD.

• add test
• split netbsd specific code to a new function