[Libreoffice-commits] core.git: comphelper/source external/firebird
Stephan Bergmann
sbergman at redhat.com
Thu Dec 28 12:51:11 UTC 2017
comphelper/source/misc/asyncnotification.cxx | 1
external/firebird/UnpackedTarball_firebird.mk | 1
external/firebird/sanitizer.patch | 28 ++
external/firebird/ubsan.patch | 255 ++++++++++++++++++++++++++
4 files changed, 284 insertions(+), 1 deletion(-)
New commits:
commit 7f15b7ae482be5994c6803b982c13cbc036a734f
Author: Stephan Bergmann <sbergman at redhat.com>
Date: Thu Dec 28 09:34:44 2017 +0100
Full UBSan in external/firebird
...after 6a312a4a3d642f0f9769df54c0ec25471c9916bd "Prepare external/firebird for
sanitizers". Doing `make check screenshot` on Linux with Clang
-fsanitize=undefined -fsanitize=local-bounds -fsanitize=nullability works now.
Patches that might be unwanted in non-UBSan builds are added to sanitizer.patch
while unproblematic ones are added to always-included ubsan.patch.
CppunitTest_dbaccess_firebird_test, e.g., shows that
comphelper::AsyncEventNotifierAutoJoin::onTerminated called base-class
osl::Thread::onTerminated (which does nothing, anyway) of an already destroyed
object, so just drop that.
Change-Id: If50f442ee6dbe590db843c38681d1c1cade8fa6a
Reviewed-on: https://gerrit.libreoffice.org/47122
Tested-by: Jenkins <ci at libreoffice.org>
Reviewed-by: Stephan Bergmann <sbergman at redhat.com>
diff --git a/comphelper/source/misc/asyncnotification.cxx b/comphelper/source/misc/asyncnotification.cxx
index ec305e4bdfce..2a2ef6b1c819 100644
--- a/comphelper/source/misc/asyncnotification.cxx
+++ b/comphelper/source/misc/asyncnotification.cxx
@@ -266,7 +266,6 @@ namespace comphelper
{
// try to delete "this"
m_xImpl->pKeepThisAlive.reset();
- return osl::Thread::onTerminated();
}
} // namespace comphelper
diff --git a/external/firebird/UnpackedTarball_firebird.mk b/external/firebird/UnpackedTarball_firebird.mk
index a055b6573696..fd780b964945 100644
--- a/external/firebird/UnpackedTarball_firebird.mk
+++ b/external/firebird/UnpackedTarball_firebird.mk
@@ -29,6 +29,7 @@ $(eval $(call gb_UnpackedTarball_add_patches,firebird,\
external/firebird/0001-Avoid-hangup-in-SS-when-error-happens-at-system-atta.patch.1 \
external/firebird/0002-Backported-fix-for-CORE-5452-Segfault-when-engine-s-.patch.1 \
external/firebird/c++17.patch \
+ external/firebird/ubsan.patch \
))
ifeq ($(OS),WNT)
diff --git a/external/firebird/sanitizer.patch b/external/firebird/sanitizer.patch
index cf3a54b266f8..e727d581e626 100644
--- a/external/firebird/sanitizer.patch
+++ b/external/firebird/sanitizer.patch
@@ -39,6 +39,17 @@
LIB_LINK_OPTIONS= $(LDFLAGS) $(THR_FLAGS) -shared
FB_DAEMON = $(BIN)/firebird$(EXEC_EXT)
+--- src/common/classes/alloc.cpp
++++ src/common/classes/alloc.cpp
+@@ -2535,7 +2535,7 @@
+ const char* myStack = &probeVar;
+ const char* thisLocation = (const char*) this;
+ ptrdiff_t distance = thisLocation - myStack;
+- fb_assert(absVal(distance) < 128 * 1024);
++ //fb_assert(absVal(distance) < 128 * 1024);
+ }
+ #endif
+
--- src/common/os/posix/mod_loader.cpp
+++ src/common/os/posix/mod_loader.cpp
@@ -88,7 +88,7 @@
@@ -50,3 +61,20 @@
if (module == NULL)
{
#ifdef DEV_BUILD
+--- src/jrd/met.epp
++++ src/jrd/met.epp
+@@ -1486,10 +1486,11 @@
+ USHORT offset = p[0] | (p[1] << 8);
+ p += 2;
+
+- const Ods::Descriptor* odsDflDesc = (Ods::Descriptor*) p;
+- p = (UCHAR*) (odsDflDesc + 1);
++ Ods::Descriptor odsDflDesc;
++ memcpy(&odsDflDesc, p, sizeof (Ods::Descriptor));
++ p = (UCHAR*) (((Ods::Descriptor*) p) + 1);
+
+- dsc desc = *odsDflDesc;
++ dsc desc = odsDflDesc;
+ desc.dsc_address = const_cast<UCHAR*>(p);
+ EVL_make_value(tdbb, &desc, &format->fmt_defaults[offset], relation->rel_pool);
+
diff --git a/external/firebird/ubsan.patch b/external/firebird/ubsan.patch
new file mode 100644
index 000000000000..8ccc1703ec9b
--- /dev/null
+++ b/external/firebird/ubsan.patch
@@ -0,0 +1,255 @@
+--- configure
++++ configure
+@@ -21349,7 +21349,7 @@
+ char a;
+ long long b;
+ };
+- exit((int)&((struct s*)0)->b);
++ exit((int)&((struct s*)1024)->b - 1024);
+ }
+ _ACEOF
+ if ac_fn_c_try_run "$LINENO"; then :
+@@ -21384,7 +21384,7 @@
+ char a;
+ double b;
+ };
+- exit((int)&((struct s*)0)->b);
++ exit((int)&((struct s*)1024)->b - 1024);
+ }
+ _ACEOF
+ if ac_fn_c_try_run "$LINENO"; then :
+--- src/common/classes/array.h
++++ src/common/classes/array.h
+@@ -149,7 +149,7 @@
+ void copyFrom(const Array<T, Storage>& source)
+ {
+ ensureCapacity(source.count, false);
+- memcpy(data, source.data, sizeof(T) * source.count);
++ if (source.count != 0) memcpy(data, source.data, sizeof(T) * source.count);
+ count = source.count;
+ }
+
+@@ -227,7 +227,7 @@
+ fb_assert(count <= FB_MAX_SIZEOF - itemsCount);
+ ensureCapacity(count + itemsCount);
+ memmove(data + index + itemsCount, data + index, sizeof(T) * (count - index));
+- memcpy(data + index, items, sizeof(T) * itemsCount);
++ if (itemsCount != 0) memcpy(data + index, items, sizeof(T) * itemsCount);
+ count += itemsCount;
+ }
+
+@@ -242,7 +242,7 @@
+ {
+ fb_assert(count <= FB_MAX_SIZEOF - itemsCount);
+ ensureCapacity(count + itemsCount);
+- memcpy(data + count, items, sizeof(T) * itemsCount);
++ if (itemsCount != 0) memcpy(data + count, items, sizeof(T) * itemsCount);
+ count += itemsCount;
+ }
+
+@@ -294,7 +294,7 @@
+ {
+ fb_assert(newCount >= count);
+ ensureCapacity(newCount);
+- memset(data + count, 0, sizeof(T) * (newCount - count));
++ if (newCount != count) memset(data + count, 0, sizeof(T) * (newCount - count));
+ count = newCount;
+ }
+
+@@ -328,7 +328,7 @@
+ {
+ fb_assert(count <= FB_MAX_SIZEOF - L.count);
+ ensureCapacity(count + L.count);
+- memcpy(data + count, L.data, sizeof(T) * L.count);
++ if (L.count != 0) memcpy(data + count, L.data, sizeof(T) * L.count);
+ count += L.count;
+ }
+
+@@ -462,7 +462,7 @@
+
+ T* newdata = static_cast<T*>
+ (this->getPool().allocate(sizeof(T) * newcapacity ALLOC_ARGS));
+- if (preserve)
++ if (preserve && count != 0)
+ memcpy(newdata, data, sizeof(T) * count);
+ freeData();
+ data = newdata;
+--- src/dsql/StmtNodes.cpp
++++ src/dsql/StmtNodes.cpp
+@@ -6643,7 +6643,7 @@
+
+ void StoreNode::genBlr(DsqlCompilerScratch* dsqlScratch)
+ {
+- const dsql_msg* message = dsqlGenDmlHeader(dsqlScratch, dsqlRse->as<RseNode>());
++ const dsql_msg* message = dsqlGenDmlHeader(dsqlScratch, dsqlRse == nullptr ? nullptr : dsqlRse->as<RseNode>());
+
+ dsqlScratch->appendUChar(statement2 ? blr_store2 : blr_store);
+ GEN_expr(dsqlScratch, dsqlRelation);
+--- src/gpre/hsh.cpp
++++ src/gpre/hsh.cpp
+@@ -232,7 +232,7 @@
+ {
+ SCHAR c;
+
+- SLONG value = 0;
++ ULONG value = 0;
+
+ while (c = *string++)
+ value = (value << 1) + UPPER(c);
+--- src/jrd/GlobalRWLock.cpp
++++ src/jrd/GlobalRWLock.cpp
+@@ -78,7 +78,7 @@
+
+ cachedLock = FB_NEW_RPT(getPool(), lockLen)
+ Lock(tdbb, lockLen, lckType, this, lockCaching ? blocking_ast_cached_lock : NULL);
+- memcpy(&cachedLock->lck_key, lockStr, lockLen);
++ if (lockLen != 0) memcpy(&cachedLock->lck_key, lockStr, lockLen);
+ }
+
+ GlobalRWLock::~GlobalRWLock()
+--- src/jrd/Optimizer.cpp
++++ src/jrd/Optimizer.cpp
+@@ -368,7 +368,7 @@
+
+ // Allocate needed indexScratches
+
+- index_desc* idx = csb_tail->csb_idx->items;
++ index_desc* idx = csb_tail->csb_idx == nullptr ? nullptr : csb_tail->csb_idx->items;
+ for (int i = 0; i < csb_tail->csb_indices; ++i, ++idx)
+ indexScratches.add(IndexScratch(p, tdbb, idx, csb_tail));
+ }
+--- src/jrd/blb.cpp
++++ src/jrd/blb.cpp
+@@ -1786,7 +1786,7 @@
+ arg.slice_base = array->arr_data;
+
+ SLONG variables[64];
+- memcpy(variables, param, MIN(sizeof(variables), param_length));
++ if (param_length != 0) memcpy(variables, param, MIN(sizeof(variables), param_length));
+
+ if (SDL_walk(tdbb->tdbb_status_vector, sdl, array->arr_data, &array_desc->arr_desc,
+ variables, slice_callback, &arg))
+--- src/jrd/btn.cpp
++++ src/jrd/btn.cpp
+@@ -387,7 +387,7 @@
+
+ put_short(pagePointer, offset);
+ pagePointer += sizeof(USHORT);
+- memmove(pagePointer, data, length);
++ if (length != 0) memmove(pagePointer, data, length);
+ pagePointer += length;
+ return pagePointer;
+ }
+@@ -622,7 +622,7 @@
+ }
+
+ // Store data
+- if (withData) {
++ if (withData && length != 0) {
+ memcpy(pagePointer, data, length);
+ }
+ pagePointer += length;
+--- src/jrd/btr.cpp
++++ src/jrd/btr.cpp
+@@ -5206,7 +5206,7 @@
+ // Push node on end in list
+ jumpNodes->add(jumpNode);
+ // Store new data in jumpKey, so a new jump node can calculate prefix
+- memcpy(jumpData + jumpNode.prefix, jumpNode.data, jumpNode.length);
++ if (jumpNode.length != 0) memcpy(jumpData + jumpNode.prefix, jumpNode.data, jumpNode.length);
+ jumpLength = jumpNode.length + jumpNode.prefix;
+
+ // Check if this could be our split point (if we need to split)
+@@ -5611,7 +5611,7 @@
+ for (size_t i = 0; i < jumpNodes->getCount(); i++, index++)
+ {
+ UCHAR* q = new_key->key_data + walkJumpNode[i].prefix;
+- memcpy(q, walkJumpNode[i].data, walkJumpNode[i].length);
++ if (walkJumpNode[i].length != 0) memcpy(q, walkJumpNode[i].data, walkJumpNode[i].length);
+ if (index == splitJumpNodeIndex)
+ {
+ jn = &walkJumpNode[i];
+@@ -5636,7 +5636,7 @@
+ const USHORT length = walkJumpNode[i].prefix + walkJumpNode[i].length;
+ UCHAR* newData = FB_NEW_POOL(*tdbb->getDefaultPool()) UCHAR[length];
+ memcpy(newData, new_key->key_data, walkJumpNode[i].prefix);
+- memcpy(newData + walkJumpNode[i].prefix, walkJumpNode[i].data,
++ if (walkJumpNode[i].length != 0) memcpy(newData + walkJumpNode[i].prefix, walkJumpNode[i].data,
+ walkJumpNode[i].length);
+ delete[] walkJumpNode[i].data;
+ walkJumpNode[i].prefix = 0;
+--- src/jrd/evl.cpp
++++ src/jrd/evl.cpp
+@@ -415,7 +415,7 @@
+ case dtype_real:
+ case dtype_sql_time:
+ case dtype_sql_date:
+- value->vlu_misc.vlu_long = *((SLONG*) from.dsc_address);
++ memcpy(&value->vlu_misc.vlu_long, from.dsc_address, sizeof (SLONG));
+ return;
+
+ case dtype_int64:
+--- src/jrd/lck.cpp
++++ src/jrd/lck.cpp
+@@ -488,7 +488,7 @@
+ break;
+ }
+
+- dbb->dbb_lock_mgr->shutdownOwner(tdbb, owner_handle_ptr);
++ LockManager::shutdownOwner(dbb->dbb_lock_mgr, tdbb, owner_handle_ptr);
+ }
+
+
+--- src/lock/lock.cpp
++++ src/lock/lock.cpp
+@@ -441,7 +441,7 @@
+ }
+
+
+-void LockManager::shutdownOwner(thread_db* tdbb, SRQ_PTR* owner_handle)
++void LockManager::shutdownOwner(LockManager* This, thread_db* tdbb, SRQ_PTR* owner_handle)
+ {
+ /**************************************
+ *
+@@ -460,8 +460,9 @@
+ if (!owner_offset)
+ return;
+
+- LockTableGuard guard(this, FB_FUNCTION, owner_offset);
++ LockTableGuard guard(This, FB_FUNCTION, owner_offset);
+
++#define SRQ_BASE ((UCHAR*) This->m_sharedMemory->getHeader())
+ own* owner = (own*) SRQ_ABS_PTR(owner_offset);
+ if (!owner->own_count)
+ return;
+@@ -472,7 +473,7 @@
+ while (owner->own_ast_count)
+ {
+ { // checkout scope
+- LockTableCheckout checkout(this, FB_FUNCTION);
++ LockTableCheckout checkout(This, FB_FUNCTION);
+ EngineCheckout cout(tdbb, FB_FUNCTION, true);
+ Thread::sleep(10);
+ }
+@@ -484,8 +485,9 @@
+ // released before destroying the lock owner. This is not strictly required,
+ // but it enforces the proper object lifetime discipline through the codebase.
+ fb_assert(SRQ_EMPTY(owner->own_requests));
++#define SRQ_BASE ((UCHAR*) m_sharedMemory->getHeader())
+
+- purge_owner(owner_offset, owner);
++ This->purge_owner(owner_offset, owner);
+
+ *owner_handle = 0;
+ }
+--- src/lock/lock_proto.h
++++ src/lock/lock_proto.h
+@@ -402,7 +402,7 @@
+ static void destroy(LockManager*);
+
+ bool initializeOwner(Firebird::CheckStatusWrapper*, LOCK_OWNER_T, UCHAR, SRQ_PTR*);
+- void shutdownOwner(thread_db*, SRQ_PTR*);
++ static void shutdownOwner(LockManager* This, thread_db*, SRQ_PTR*);
+
+ SRQ_PTR enqueue(thread_db*, Firebird::CheckStatusWrapper*, SRQ_PTR, const USHORT,
+ const UCHAR*, const USHORT, UCHAR, lock_ast_t, void*, SINT64, SSHORT, SRQ_PTR);
More information about the Libreoffice-commits
mailing list