Created attachment 28731 [details] Source file A source file which contains call_once will compile without any errors or warnings, even when libpthread is not linked. At run time, an exception will be thrown with a very unhelpful message. The attached source file demonstrates the problem. $ g++ -Wall -Wextra -std=c++0x callonce.cpp $ ./a.out terminate called after throwing an instance of 'std::system_error' what(): Unknown error -1 Aborted (core dumped) Linking libpthread will remove the problem. $ g++ -Wall -Wextra -std=c++0x callonce.cpp -lpthread $ ./a.out $ In the actual use case, call_once was called in a library that could be used in both multithreaded and singlethreaded code. That made it a hard to trace bug. $ g++ -v Using built-in specs. COLLECT_GCC=g++ COLLECT_LTO_WRAPPER=/usr/lib/gcc/x86_64-linux-gnu/4.6/lto-wrapper Target: x86_64-linux-gnu Configured with: ../src/configure -v --with-pkgversion='Ubuntu/Linaro 4.6.3-1ubuntu5' --with-bugurl=file:///usr/share/doc/gcc-4.6/README.Bugs --enable-languages=c,c++,fortran,objc,obj-c++ --prefix=/usr --program-suffix=-4.6 --enable-shared --enable-linker-build-id --with-system-zlib --libexecdir=/usr/lib --without-included-gettext --enable-threads=posix --with-gxx-include-dir=/usr/include/c++/4.6 --libdir=/usr/lib --enable-nls --with-sysroot=/ --enable-clocale=gnu --enable-libstdcxx-debug --enable-libstdcxx-time=yes --enable-gnu-unique-object --enable-plugin --enable-objc-gc --disable-werror --with-arch-32=i686 --with-tune=generic --enable-checking=release --build=x86_64-linux-gnu --host=x86_64-linux-gnu --target=x86_64-linux-gnu Thread model: posix gcc version 4.6.3 (Ubuntu/Linaro 4.6.3-1ubuntu5)
(In reply to comment #0) > A source file which contains call_once will compile without any errors or > warnings, even when libpthread is not linked. Obviously this is because compiling and linking are separate steps. The compiler front-end doesn't know in advance whether the linker will be invoked correctly. It might be possible to do something similar to PR 52681 to improve the exception's message.
Library anyway.
related to PR 58929
clang's libc++.so.1 avoids the issue because it is linked with libpthread.so.0: % ldd /usr/lib/libc++.so.1 linux-vdso.so.1 (0x00007ffc1618f000) libpthread.so.0 => /lib/libpthread.so.0 (0x00007fe0f4fd7000) libc.so.6 => /lib/libc.so.6 (0x00007fe0f4c3a000) libm.so.6 => /lib/libm.so.6 (0x00007fe0f499f000) librt.so.1 => /lib/librt.so.1 (0x00007fe0f4797000) libgcc_s.so.1 => /lib/libgcc_s.so.1 (0x00007fe0f477e000) /lib64/ld-linux-x86-64.so.2 (0x00005575a195e000) libstc++ isn't: % ldd /usr/lib/gcc/x86_64-pc-linux-gnu/7.0.0/libstdc++.so.6 linux-vdso.so.1 (0x00007ffd14540000) libm.so.6 => /lib/libm.so.6 (0x00007f73bec87000) libc.so.6 => /lib/libc.so.6 (0x00007f73be8ea000) /lib64/ld-linux-x86-64.so.2 (0x00005594a5a06000) libgcc_s.so.1 => /lib/libgcc_s.so.1 (0x00007f73be8d1000)
(In reply to Markus Trippelsdorf from comment #4) > clang's libc++.so.1 avoids the issue because it is linked with > libpthread.so.0: > [...] > libstc++ isn't: Which is by design and a very important property to keep. I was discussing a similar issue on IRC the other day and came to the conclusion that we should probably just call pthread_once() directly when the thread model is posix. Because that call is in a header not in libstdc++.so it won't make libstdc++.so depend on libpthread.so, but users that call std::call_once will depend on libpthread.so and users that don't call std::call_once won't depend on it. Which is what we want.
*** Bug 81517 has been marked as a duplicate of this bug. ***
The same test program will fail in the same way, if compiled with -flto, even if -pthread is used: $ g++-9 -Wall -Wextra -std=c++17 callonce.cpp -pthread $ ./a.out but $ g++-9 -Wall -Wextra -std=c++17 callonce.cpp -pthread -flto $ ./a.out terminate called after throwing an instance of 'std::system_error' what(): Unknown error -1 Aborted (core dumped) Possibly because the LTO pass discards the dependency on libpthread.so: $ ldd a.out linux-vdso.so.1 (0x00007ffc09b46000) libc.so.6 => /lib/x86_64-linux-gnu/libc.so.6 (0x00007f420535e000) libstdc++.so.6 => /usr/lib/x86_64-linux-gnu/libstdc++.so.6 (0x00007f4204f7e000) /lib64/ld-linux-x86-64.so.2 (0x00007f4205952000) libm.so.6 => /lib/x86_64-linux-gnu/libm.so.6 (0x00007f4204be0000) libgcc_s.so.1 => /lib/x86_64-linux-gnu/libgcc_s.so.1 (0x00007f42049c8000)
Any progress on that issue? Looks like I've hit this issu in protobuf https://github.com/protocolbuffers/protobuf/issues/7092
*** Bug 97485 has been marked as a duplicate of this bug. ***
How about something as simple as: --- a/libgcc/gthr-posix.h +++ b/libgcc/gthr-posix.h @@ -697,7 +697,12 @@ static inline int __gthread_once (__gthread_once_t *__once, void (*__func) (void)) { if (__gthread_active_p ()) - return __gthrw_(pthread_once) (__once, __func); + /* + * Call non-weak form of a symbol to force linker + * resolve real implementation of pthread_once even + * for as-needed case: https://gcc.gnu.org/PR55394. + */ + return pthread_once (__once, __func); else return -1; } Superficial test fixes "-flto -Wl,--as-needed" case above for me. The fix leaves '__gthread_active_p ()' to probe weak symbols but calls non-weak form of pthread_once(). If it's a plausible path I can have a pass through all static inline functions, and convert weak calls to non-weak calls where glibc-2.32 does not define a symbol in libc.so.6.
It's not plausible because it doesn't work for non-pthreads targets where gthr-default.h is not gthr-posix.h We can't use pthread_once anyway, see PR 66146, so I'm rewriting it entirely in terms of either futexes or condition variables.
Aha, makes sense. My hack did not survive bootstrap anyway as libgcc.a started referring pthread_once() as well.
The master branch has been updated by Jonathan Wakely <redi@gcc.gnu.org>: https://gcc.gnu.org/g:93e79ed391b9c636f087e6eb7e70f14963cd10ad commit r11-4691-g93e79ed391b9c636f087e6eb7e70f14963cd10ad Author: Jonathan Wakely <jwakely@redhat.com> Date: Tue Nov 3 18:44:32 2020 +0000 libstdc++: Rewrite std::call_once to use futexes [PR 66146] The current implementation of std::call_once uses pthread_once, which only meets the C++ requirements when compiled with support for exceptions. For most glibc targets and all non-glibc targets, pthread_once does not work correctly if the init_routine exits via an exception. The pthread_once_t object is left in the "active" state, and any later attempts to run another init_routine will block forever. This change makes std::call_once work correctly for Linux targets, by replacing the use of pthread_once with a futex, based on the code from __cxa_guard_acquire. For both glibc and musl, the Linux implementation of pthread_once is already based on futexes, and pthread_once_t is just a typedef for int, so this change does not alter the layout of std::once_flag. By choosing the values for the int appropriately, the new code is even ABI compatible. Code that calls the old implementation of std::call_once will use pthread_once to manipulate the int, while new code will use the new std::once_flag members to manipulate it, but they should interoperate correctly. In both cases, the int is initially zero, has the lowest bit set when there is an active execution, and equals 2 after a successful returning execution. The difference with the new code is that exceptional exceptions are correctly detected and the int is reset to zero. The __cxa_guard_acquire code (and musl's pthread_once) use an additional state to say there are other threads waiting. This allows the futex wake syscall to be skipped if there is no contention. Glibc doesn't use a waiter bit, so we have to unconditionally issue the wake in order to be compatible with code calling the old std::call_once that uses Glibc's pthread_once. If we know that we're using musl (and musl's pthread_once doesn't change) it would be possible to set a waiting state and check for it in std::once_flag::_M_finish(bool), but this patch doesn't do that. This doesn't fix the bug for non-linux targets. A similar approach could be used for targets where we know the definition of pthread_once_t is a mutex and an integer. We could make once_flag._M_activate() use pthread_mutex_lock on the mutex member within the pthread_once_t, and then only set the integer if the execution finishes, and then unlock the mutex. That would require careful study of each target's pthread_once implementation and that work is left for a later date. This also fixes PR 55394 because pthread_once is no longer needed, and PR 84323 because the fast path is now just an atomic load. As a consequence of the new implementation that doesn't use pthread_once, we can also make std::call_once work for targets with no gthreads support. The code for the single-threaded implementation follows the same methods as on Linux, but with no need for atomics or futexes. libstdc++-v3/ChangeLog: PR libstdc++/55394 PR libstdc++/66146 PR libstdc++/84323 * config/abi/pre/gnu.ver (GLIBCXX_3.4.29): Add new symbols. * include/std/mutex [!_GLIBCXX_HAS_GTHREADS] (once_flag): Define even when gthreads is not supported. (once_flag::_M_once) [_GLIBCXX_HAVE_LINUX_FUTEX]: Change type from __gthread_once_t to int. (once_flag::_M_passive(), once_flag::_M_activate()) (once_flag::_M_finish(bool), once_flag::_Active_execution): Define new members for futex and non-threaded implementation. [_GLIBCXX_HAS_GTHREADS] (once_flag::_Prepare_execution): New RAII helper type. (call_once): Use new members of once_flag. * src/c++11/mutex.cc (std::once_flag::_M_activate): Define. (std::once_flag::_M_finish): Define. * testsuite/30_threads/call_once/39909.cc: Do not require gthreads. * testsuite/30_threads/call_once/49668.cc: Likewise. * testsuite/30_threads/call_once/60497.cc: Likewise. * testsuite/30_threads/call_once/call_once1.cc: Likewise. * testsuite/30_threads/call_once/dr2442.cc: Likewise. * testsuite/30_threads/call_once/once_flag.cc: Add test for constexpr constructor. * testsuite/30_threads/call_once/66146.cc: New test. * testsuite/30_threads/call_once/constexpr.cc: Removed. * testsuite/30_threads/once_flag/cons/constexpr.cc: Removed.
Fixed for GCC 11.