a0072d0a31
add some more deadlock suppressions Change-Id: I457e66275921080f9f12c3ab2f2fbbb6168ad108 Reviewed-on: https://gerrit.libreoffice.org/c/core/+/146345 Tested-by: Noel Grandin <noel.grandin@collabora.co.uk> Reviewed-by: Noel Grandin <noel.grandin@collabora.co.uk>
68 lines
2.2 KiB
Text
68 lines
2.2 KiB
Text
#
|
|
# This file is part of the LibreOffice project.
|
|
#
|
|
# This Source Code Form is subject to the terms of the Mozilla Public
|
|
# License, v. 2.0. If a copy of the MPL was not distributed with this
|
|
# file, You can obtain one at http://mozilla.org/MPL/2.0/.
|
|
#
|
|
|
|
# Use via environment variable TSAN_OPTIONS=suppressions=.../tsan-suppress.txt
|
|
|
|
# There looks to be a lock ordering problem here, but I can't see how it could
|
|
# actually be hit in practice.
|
|
deadlock:cppuhelper::ServiceManager::disposing()
|
|
deadlock:cppuhelper::ServiceManager::loadImplementation
|
|
deadlock:AffineBridge::v_callInto_v
|
|
|
|
# Ignore stuff in external DBUS library.
|
|
# Some kind of dbus lock issue when we call it from psp::CUPSManager.
|
|
deadlock:_dbus_lock
|
|
|
|
# inside an assert so I don't care.
|
|
race:cppu::OWeakConnectionPoint::acquire
|
|
race:AffineBridge::v_enter
|
|
race:__vsnprintf_chk
|
|
|
|
# right now, I'm not interested in deadlocks at all, too many false+
|
|
deadlock:
|
|
|
|
# This is checking SAL_STRING_IS_STATIC, which is safe because that is written at compile time.
|
|
# Depending on the optimisation/debug level, the call stack may have different leaves, hence
|
|
# specifying this in more than one way.
|
|
race:rtl::str::acquire<_rtl_uString>
|
|
race:rtl::str::release<_rtl_uString>
|
|
race:rtl_uString_acquire
|
|
race:rtl_uString_release
|
|
race:rtl_uString_assign
|
|
|
|
# I've convinced myself this is a false+, caused by ping-ponging the buffer between two
|
|
# threads, but I might be wrong.
|
|
race:XBufferedThreadedStream::getNextBlock
|
|
|
|
# I am not interested in stuff the embedded JVM does.
|
|
race:libjvm.so
|
|
|
|
# I think this is OK, because at this point we are doing
|
|
# if (nRefCount > 1)
|
|
# and we know from our callers that the refcount must be at least one
|
|
# so there is no failure mode.
|
|
race:ireallocSequence
|
|
|
|
# TODO There appears to be a race here, initialising the
|
|
# ::com::sun::star::uno::Sequence< T >::s_pType
|
|
# field. But no idea at all how to fix it.
|
|
race:cppu::getTypeFavourUnsigned
|
|
|
|
|
|
# This is all inside GIO/Glib, no idea what it is doing.
|
|
#
|
|
race:slab_allocator_alloc_chunk
|
|
race:g_source_destroy_internal
|
|
race:g_source_unref_internal
|
|
race:g_task_finalize
|
|
race:g_socket_send_message_with_timeout
|
|
race:g_idle_source_new
|
|
race:g_slice_alloc0
|
|
|
|
# Python build, lock order inversion
|
|
deadlock:take_gil
|