[rtsan] Intercept aligned_alloc on all versions of OSX if available on build machine #112780

cjappl · 2024-10-17T21:41:15Z

This follows the pattern seen here, where tsan uses libdispatch, even if the functions aren't available in our minimum version.

llvm-project/compiler-rt/lib/tsan/rtl/tsan_interceptors_libdispatch.cpp

Lines 226 to 247 in f7669ba

    
           // dispatch_async_and_wait() and friends were introduced in macOS 10.14. 
        
           // Linking of these interceptors fails when using an older SDK. 
        
           #if !SANITIZER_APPLE || defined(__MAC_10_14) 
        
           // macOS 10.14 is greater than our minimal deployment target.  To ensure we 
        
           // generate a weak reference so the TSan dylib continues to work on older 
        
           // systems, we need to forward declare the intercepted functions as "weak 
        
           // imports".   Note that this file is multi-platform, so we cannot include the 
        
           // actual header file (#include <dispatch/dispatch.h>). 
        
           SANITIZER_WEAK_IMPORT void dispatch_async_and_wait( 
        
               dispatch_queue_t queue, DISPATCH_NOESCAPE dispatch_block_t block); 
        
           SANITIZER_WEAK_IMPORT void dispatch_async_and_wait_f( 
        
               dispatch_queue_t queue, void *context, dispatch_function_t work); 
        
           SANITIZER_WEAK_IMPORT void dispatch_barrier_async_and_wait( 
        
               dispatch_queue_t queue, DISPATCH_NOESCAPE dispatch_block_t block); 
        
           SANITIZER_WEAK_IMPORT void dispatch_barrier_async_and_wait_f( 
        
               dispatch_queue_t queue, void *context, dispatch_function_t work); 
        
           DISPATCH_INTERCEPT_SYNC_F(dispatch_async_and_wait_f, false) 
        
           DISPATCH_INTERCEPT_SYNC_B(dispatch_async_and_wait, false) 
        
           DISPATCH_INTERCEPT_SYNC_F(dispatch_barrier_async_and_wait_f, true) 
        
           DISPATCH_INTERCEPT_SYNC_B(dispatch_barrier_async_and_wait, true) 
        
           #endif

I think it would be good to propagate this change to the other sanitizers, as currently, no sanitizer actually intercepts aligned alloc on mac, stemming from:

llvm-project/compiler-rt/lib/sanitizer_common/sanitizer_platform_interceptors.h

Line 497 in f7669ba

#define SANITIZER_INTERCEPT_ALIGNED_ALLOC (!SI_MAC)

Ideally, if this pattern looked good, we could add it to sanitizer_malloc_mac.inc, and the other sanitizers would have this enabled.

llvmbot · 2024-10-17T21:41:49Z

@llvm/pr-subscribers-compiler-rt-sanitizer

Author: Chris Apple (cjappl)

Changes

This follows the pattern seen here, where tsan uses libdispatch, even if the functions aren't available in our minimum version.

llvm-project/compiler-rt/lib/tsan/rtl/tsan_interceptors_libdispatch.cpp

Lines 226 to 247 in f7669ba

    
           // dispatch_async_and_wait() and friends were introduced in macOS 10.14. 
        
           // Linking of these interceptors fails when using an older SDK. 
        
           #if !SANITIZER_APPLE || defined(__MAC_10_14) 
        
           // macOS 10.14 is greater than our minimal deployment target.  To ensure we 
        
           // generate a weak reference so the TSan dylib continues to work on older 
        
           // systems, we need to forward declare the intercepted functions as "weak 
        
           // imports".   Note that this file is multi-platform, so we cannot include the 
        
           // actual header file (#include <dispatch/dispatch.h>). 
        
           SANITIZER_WEAK_IMPORT void dispatch_async_and_wait( 
        
               dispatch_queue_t queue, DISPATCH_NOESCAPE dispatch_block_t block); 
        
           SANITIZER_WEAK_IMPORT void dispatch_async_and_wait_f( 
        
               dispatch_queue_t queue, void *context, dispatch_function_t work); 
        
           SANITIZER_WEAK_IMPORT void dispatch_barrier_async_and_wait( 
        
               dispatch_queue_t queue, DISPATCH_NOESCAPE dispatch_block_t block); 
        
           SANITIZER_WEAK_IMPORT void dispatch_barrier_async_and_wait_f( 
        
               dispatch_queue_t queue, void *context, dispatch_function_t work); 
        
           DISPATCH_INTERCEPT_SYNC_F(dispatch_async_and_wait_f, false) 
        
           DISPATCH_INTERCEPT_SYNC_B(dispatch_async_and_wait, false) 
        
           DISPATCH_INTERCEPT_SYNC_F(dispatch_barrier_async_and_wait_f, true) 
        
           DISPATCH_INTERCEPT_SYNC_B(dispatch_barrier_async_and_wait, true) 
        
           #endif

I think it would be good to propagate this change to the other sanitizers, as currently, no sanitizer actually intercepts aligned alloc on mac, stemming from:

llvm-project/compiler-rt/lib/sanitizer_common/sanitizer_platform_interceptors.h

Line 497 in f7669ba

#define SANITIZER_INTERCEPT_ALIGNED_ALLOC (!SI_MAC)

Ideally, if this pattern looked good, we could add it to sanitizer_malloc_mac.inc, and the other sanitizers would have this enabled.

Full diff: https://github.com/llvm/llvm-project/pull/112780.diff

2 Files Affected:

(modified) compiler-rt/lib/rtsan/rtsan_interceptors.cpp (+13-1)
(modified) compiler-rt/lib/rtsan/tests/rtsan_test_interceptors.cpp (+9-5)

diff --git a/compiler-rt/lib/rtsan/rtsan_interceptors.cpp b/compiler-rt/lib/rtsan/rtsan_interceptors.cpp
index 63b0ca28a1f409..7e2e9c71786c06 100644
--- a/compiler-rt/lib/rtsan/rtsan_interceptors.cpp
+++ b/compiler-rt/lib/rtsan/rtsan_interceptors.cpp
@@ -461,7 +461,17 @@ INTERCEPTOR(void *, valloc, SIZE_T size) {
   return REAL(valloc)(size);
 }
 
-#if SANITIZER_INTERCEPT_ALIGNED_ALLOC
+// aligned_alloc was introduced in OSX 10.15
+// Linking will fail when using an older SDK
+#if defined(__MAC_10_15)
+// macOS 10.15 is greater than our minimal deployment target.  To ensure we
+// generate a weak reference so the dylib continues to work on older
+// systems, we need to forward declare the intercepted function as "weak
+// imports".
+SANITIZER_WEAK_IMPORT void *aligned_alloc(SIZE_T __alignment, SIZE_T __size);
+#endif // defined(__MAC_10_15)
+
+#if SANITIZER_INTERCEPT_ALIGNED_ALLOC || SANITIZER_APPLE
 INTERCEPTOR(void *, aligned_alloc, SIZE_T alignment, SIZE_T size) {
   __rtsan_notify_intercepted_call("aligned_alloc");
   return REAL(aligned_alloc)(alignment, size);
@@ -471,6 +481,8 @@ INTERCEPTOR(void *, aligned_alloc, SIZE_T alignment, SIZE_T size) {
 #define RTSAN_MAYBE_INTERCEPT_ALIGNED_ALLOC
 #endif
 
+#undef RTSAN_DARWIN_INTERCEPT_ALIGNED_ALLOC
+
 INTERCEPTOR(int, posix_memalign, void **memptr, size_t alignment, size_t size) {
   __rtsan_notify_intercepted_call("posix_memalign");
   return REAL(posix_memalign)(memptr, alignment, size);
diff --git a/compiler-rt/lib/rtsan/tests/rtsan_test_interceptors.cpp b/compiler-rt/lib/rtsan/tests/rtsan_test_interceptors.cpp
index f7a281f13e2ff6..fb227a1d1bc05a 100644
--- a/compiler-rt/lib/rtsan/tests/rtsan_test_interceptors.cpp
+++ b/compiler-rt/lib/rtsan/tests/rtsan_test_interceptors.cpp
@@ -120,13 +120,17 @@ TEST(TestRtsanInterceptors, VallocDiesWhenRealtime) {
   ExpectNonRealtimeSurvival(Func);
 }
 
-#if SANITIZER_INTERCEPT_ALIGNED_ALLOC
 TEST(TestRtsanInterceptors, AlignedAllocDiesWhenRealtime) {
-  auto Func = []() { EXPECT_NE(nullptr, aligned_alloc(16, 32)); };
-  ExpectRealtimeDeath(Func, "aligned_alloc");
-  ExpectNonRealtimeSurvival(Func);
-}
+#if SANITIZER_APPLE
+  if (__builtin_available(macOS 10.15, *)) {
+#endif // SANITIZER_APPLE
+    auto Func = []() { EXPECT_NE(nullptr, aligned_alloc(16, 32)); };
+    ExpectRealtimeDeath(Func, "aligned_alloc");
+    ExpectNonRealtimeSurvival(Func);
+#if SANITIZER_APPLE
+  }
 #endif
+}
 
 // free_sized and free_aligned_sized (both C23) are not yet supported
 TEST(TestRtsanInterceptors, FreeDiesWhenRealtime) {

cjappl · 2024-10-17T21:42:18Z

@yln specifically would love your eyes on this, as you implemented this pattern for TSAN. Does this seem reasonable to you? Any pitfalls I should worry about?

cjappl · 2024-10-17T21:43:04Z

compiler-rt/lib/rtsan/tests/rtsan_test_interceptors.cpp

-  ExpectNonRealtimeSurvival(Func);
-}
+#if SANITIZER_APPLE
+  if (__builtin_available(macOS 10.15, *)) {


This is the only place in compiler-rt that __builtin_available is used, is there a more preferable way to do this? If I don't, I get a warning that suggests this:

warning: 'aligned_alloc' is only available on macOS 10.15 or newer [-Wu nguarded-availability-new] ... _malloc.h:65:7: note: 'aligned_alloc' has been marked as being introduced in macOS 10.15 here, but the deployment target is macOS 10.13.0 65 | void *aligned_alloc(size_t __alignment, size_t __size) __result_use_check __alloc_align(1) __alloc_size(2) _MALLOC_TYPED(malloc_type_aligned_alloc, 2) __OSX_AVAILABLE(10.15) __IOS_AVAILABLE(13.0) __TVOS_AVAILABLE(13.0) __WATCHOS_AVAILABLE(6.0); ... note: enclose 'aligned_alloc' in a __builtin_available check to silence this warning

In my catch-up reading about this issue I found this documentation page, which suggests comparing the weak function pointer to NULL as a check. Not sure if this would be preferable here, but thought I'd drop it here just in case you hadn't seen it already 👍

Great find on those docs, I'll spend some time digging through them.

Taking the very first naive crack at it, it appears to have similar errors (please enclose in __builtin_available).

That was repro'd applying this patch:

diff --git a/compiler-rt/lib/rtsan/tests/rtsan_test_interceptors.cpp b/compiler-rt/lib/rtsan/tests/rtsan_test_interceptors.cpp index fb227a1d1bc0..ad5c52fe7afc 100644 --- a/compiler-rt/lib/rtsan/tests/rtsan_test_interceptors.cpp +++ b/compiler-rt/lib/rtsan/tests/rtsan_test_interceptors.cpp @@ -120,9 +120,11 @@ TEST(TestRtsanInterceptors, VallocDiesWhenRealtime) { ExpectNonRealtimeSurvival(Func); } +extern void* aligned_alloc(size_t, size_t) __attribute__((weak_import)); + TEST(TestRtsanInterceptors, AlignedAllocDiesWhenRealtime) { #if SANITIZER_APPLE - if (__builtin_available(macOS 10.15, *)) { + if (aligned_alloc != nullptr) #endif // SANITIZER_APPLE auto Func = []() { EXPECT_NE(nullptr, aligned_alloc(16, 32)); }; ExpectRealtimeDeath(Func, "aligned_alloc");

Also, doing some additional digging here, it seems like this __builtin_available is used in almost all subprojects of LLVM except compiler-rt, so there is reasonable support for it in the superproject at least.

Nice, in that case it sounds like __builtin_available is the recommended way forward, then?

All signs seem to be pointing that way. I definitely worry (since the std::variant debacle) about using something in compiler-rt that hasn't been used before, but seeing as it is only in a test file it seems less risky.

Also to be clear, this test passes on my mac that has aligned_alloc available, so this all works as intended :)

compiler-rt/lib/rtsan/rtsan_interceptors.cpp

vitalybuka · 2024-10-22T17:17:42Z

compiler-rt/lib/rtsan/rtsan_interceptors.cpp

@@ -461,6 +461,19 @@ INTERCEPTOR(void *, valloc, SIZE_T size) {
  return REAL(valloc)(size);
 }

+// aligned_alloc was introduced in OSX 10.15
+// Linking will fail when using an older SDK
+#if SANITIZER_APPLE && defined(__MAC_10_15)


Can we just put this into sanitizer_platform_interceptors.h ?

I moved it up in the latest commit. I initially did not put it on the top level to avoid any possible breakages with other sanitizers.

After looking, it seems like the only sanitizer that interacts with aligned_alloc outside of a *_linux.cpp file is lsan, and I feel reasonably confident this should work for that sanitizer with the weak import defined.

…n the build machine

compiler-rt/lib/sanitizer_common/sanitizer_platform_interceptors.h

cjappl · 2024-10-24T14:37:41Z

Weekly reviewer ping for outstanding.

Specifically would love @yln 's thoughts.

I'll merge this early next week if no new info

wrotki · 2024-10-28T23:43:58Z

Unfortunately this breaks the build on our (automerger) bots, which have -mmacosx-version-min=10.13 and also -Werror=unguarded-availability-new . I was thinking about patching it via wrapping in __builtin_available check (which I believe is the right one to use, as it should match the -mmacosx-version-min ) - but can't actually think of a quick fix, due to interceptors being defined via C macros.

Can I ask you to revert this for a while, so we can come up with a right remedy?

cjappl · 2024-10-28T23:48:41Z

Unfortunately this breaks the build on our (automerger) bots, which have -mmacosx-version-min=10.13 and also -Werror=unguarded-availability-new . I was thinking about patching it via wrapping in __builtin_available check (which I believe is the right one to use, as it should match the -mmacosx-version-min ) - but can't actually think of a quick fix, due to interceptors being defined via C macros.

Can I ask you to revert this for a while, so we can come up with a right remedy?

Sorry for the breakage @wrotki !

Can you share the exact error message that is the problem? I was hopeful that the __builtin_available I used in the unit test was enough to cover us.

After I have a look at that, happy to revert if there is no clear solution.

wrotki · 2024-10-28T23:51:06Z

Here goes:

llvm-project/compiler-rt/lib/rtsan/rtsan_interceptors_posix.cpp:475:21: error: 'aligned_alloc' is only available on macOS 10.15 or newer [-Werror,-Wunguarded-availability-new] 475 | INTERCEPTOR(void *, aligned_alloc, SIZE_T alignment, SIZE_T size) {

cjappl · 2024-10-28T23:52:00Z

Ah, what a pain.

Ok, I'll revert and ping you in the reversion. I would love your help in trying to get a fix that works, as I don't know how I would repro this on my machine.

wrotki · 2024-10-28T23:52:27Z

Great, thanks a lot. I'll be watching here. The repro - I guess you could specify these clang options: -mmacosx-version-min=10.13 -Werror=unguarded-availability-new in your local build to get the error

…ilable on build machine (llvm#112780)" This reverts commit 97fb21a.

…ilable on build machine (#112780)" (#113982) This reverts commit 97fb21a. Due to issue brought up in #112780 > Unfortunately this breaks the build on our (automerger) bots, which have -mmacosx-version-min=10.13 and also -Werror=unguarded-availability-new . I was thinking about patching it via wrapping in __builtin_available check (which I believe is the right one to use, as it should match the -mmacosx-version-min ) - but can't actually think of a quick fix, due to interceptors being defined via C macros. > llvm-project/compiler-rt/lib/rtsan/rtsan_interceptors_posix.cpp:475:21: error: 'aligned_alloc' is only available on macOS 10.15 or newer [-Werror,-Wunguarded-availability-new] 475 | INTERCEPTOR(void *, aligned_alloc, SIZE_T alignment, SIZE_T size) {

cjappl · 2024-10-28T23:56:33Z

That revert has been submitted. I'll try and get a more substantial fix for you to try, let me know if you have any breakthroughs

cjappl · 2024-10-29T13:46:42Z

@wrotki - a few thoughts here, wondering if I can get your input.

First, I can't repro this issue still. Strangely I do get the warning in the unit tests (which is why I used __builtin_available in the tests) but not in the source. It appears I'm just using the default sanitizer min OSX version (10.13)

Second, what if we just disabled this warning for aligned_alloc? (took this original patch, and additionally)

diff --git a/compiler-rt/lib/rtsan/rtsan_interceptors_posix.cpp b/compiler-rt/lib/rtsan/rtsan_interceptors_posix.cpp
index 890d6c11c407..7be1f07733f0 100644
--- a/compiler-rt/lib/rtsan/rtsan_interceptors_posix.cpp
+++ b/compiler-rt/lib/rtsan/rtsan_interceptors_posix.cpp
@@ -464,10 +464,13 @@ INTERCEPTOR(void *, valloc, SIZE_T size) {
 }

 #if SANITIZER_INTERCEPT_ALIGNED_ALLOC
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wunguarded-availability-new"
 INTERCEPTOR(void *, aligned_alloc, SIZE_T alignment, SIZE_T size) {
   __rtsan_notify_intercepted_call("aligned_alloc");
   return REAL(aligned_alloc)(alignment, size);
 }
+#pragma clang diagnostic pop
 #define RTSAN_MAYBE_INTERCEPT_ALIGNED_ALLOC INTERCEPT_FUNCTION(aligned_alloc)
 #else
 #define RTSAN_MAYBE_INTERCEPT_ALIGNED_ALLOC

Why is this OK?

(Theory)
We know that it will never be called on systems that do not have aligned_alloc defined because the client OSX won't provide it. This function is actually guarded on OS's lower than 10.15 because aligned_alloc declaration won't exist on that version. There will be no way to call this function from code.

Third (as a completely different thought)

This approach was cribbed from the tsan approach, which pre-declares some libdispatch API only available in 10.14. However, it seems as if your build bot has been happy with it. So far, I haven't been able to see the difference between my approach and the tsan approach, maybe there is some clue there you can see?

The code I based this off of is here:

llvm-project/compiler-rt/lib/tsan/rtl/tsan_interceptors_libdispatch.cpp

Lines 226 to 247 in f7669ba

    
           // dispatch_async_and_wait() and friends were introduced in macOS 10.14. 
        
           // Linking of these interceptors fails when using an older SDK. 
        
           #if !SANITIZER_APPLE || defined(__MAC_10_14) 
        
           // macOS 10.14 is greater than our minimal deployment target.  To ensure we 
        
           // generate a weak reference so the TSan dylib continues to work on older 
        
           // systems, we need to forward declare the intercepted functions as "weak 
        
           // imports".   Note that this file is multi-platform, so we cannot include the 
        
           // actual header file (#include <dispatch/dispatch.h>). 
        
           SANITIZER_WEAK_IMPORT void dispatch_async_and_wait( 
        
               dispatch_queue_t queue, DISPATCH_NOESCAPE dispatch_block_t block); 
        
           SANITIZER_WEAK_IMPORT void dispatch_async_and_wait_f( 
        
               dispatch_queue_t queue, void *context, dispatch_function_t work); 
        
           SANITIZER_WEAK_IMPORT void dispatch_barrier_async_and_wait( 
        
               dispatch_queue_t queue, DISPATCH_NOESCAPE dispatch_block_t block); 
        
           SANITIZER_WEAK_IMPORT void dispatch_barrier_async_and_wait_f( 
        
               dispatch_queue_t queue, void *context, dispatch_function_t work); 
        
           DISPATCH_INTERCEPT_SYNC_F(dispatch_async_and_wait_f, false) 
        
           DISPATCH_INTERCEPT_SYNC_B(dispatch_async_and_wait, false) 
        
           DISPATCH_INTERCEPT_SYNC_F(dispatch_barrier_async_and_wait_f, true) 
        
           DISPATCH_INTERCEPT_SYNC_B(dispatch_barrier_async_and_wait, true) 
        
           #endif

wrotki · 2024-10-29T21:00:29Z

Ah, yes, that's a good idea, silence the warning so it won't trigger build failure on the bot which upgrades to error. I applied this to my local llvm-project clone and the build worked. I agree that this will never work on 10.13, and fine with it.

If you want to reproduce yourself, you might want to clone llvm-project from Apple fork on Github: [email protected]:swiftlang/llvm-project.git

cjappl · 2024-10-29T23:38:07Z

Sounds good, I'm going to momentarily put up a PR re-landing this. I'll add you as a reviewer. Thanks for running it on your fork and ensuring it works!!

…ailable on build machine (llvm#112780)" (llvm#113982) This reverts commit 7c55426.

…n build machine (llvm#112780)

…ilable on build machine (llvm#112780)" (llvm#113982) This reverts commit 97fb21a. Due to issue brought up in llvm#112780 > Unfortunately this breaks the build on our (automerger) bots, which have -mmacosx-version-min=10.13 and also -Werror=unguarded-availability-new . I was thinking about patching it via wrapping in __builtin_available check (which I believe is the right one to use, as it should match the -mmacosx-version-min ) - but can't actually think of a quick fix, due to interceptors being defined via C macros. > llvm-project/compiler-rt/lib/rtsan/rtsan_interceptors_posix.cpp:475:21: error: 'aligned_alloc' is only available on macOS 10.15 or newer [-Werror,-Wunguarded-availability-new] 475 | INTERCEPTOR(void *, aligned_alloc, SIZE_T alignment, SIZE_T size) {

llvmbot added compiler-rt compiler-rt:sanitizer labels Oct 17, 2024

cjappl requested review from fmayer, MaskRay, yln and davidtrevelyan October 17, 2024 21:41

cjappl force-pushed the aligned_alloc_all_plats branch from 9ac834e to d5b090a Compare October 17, 2024 21:41

cjappl commented Oct 17, 2024

View reviewed changes

cjappl requested a review from vitalybuka October 17, 2024 21:45

fmayer reviewed Oct 18, 2024

View reviewed changes

compiler-rt/lib/rtsan/rtsan_interceptors.cpp Outdated Show resolved Hide resolved

vitalybuka reviewed Oct 22, 2024

View reviewed changes

cjappl added 4 commits October 22, 2024 10:25

[rtsan] Intercept aligned_alloc on all versions of OSX if available o…

0a01c4c

…n the build machine

Refactor builtin available to macro

4c83de3

[PR] fmayer - redef SANITIZER_ALIGNED_ALLOC as appopriate

f2c3523

[PR] vitalybuka - move aligned_alloc def up to all interceptors

1e127da

cjappl force-pushed the aligned_alloc_all_plats branch from 1e261bd to 1e127da Compare October 22, 2024 18:05

vitalybuka reviewed Oct 22, 2024

View reviewed changes

compiler-rt/lib/sanitizer_common/sanitizer_platform_interceptors.h Show resolved Hide resolved

vitalybuka approved these changes Oct 22, 2024

View reviewed changes

[PR] vitalybuka - clean up header

55e1021

cjappl merged commit 97fb21a into llvm:main Oct 28, 2024
7 checks passed

cjappl deleted the aligned_alloc_all_plats branch October 28, 2024 17:33

cjappl added a commit to cjappl/llvm-project that referenced this pull request Oct 28, 2024

Revert "[rtsan] Intercept aligned_alloc on all versions of OSX if ava…

9670688

…ilable on build machine (llvm#112780)" This reverts commit 97fb21a.

cjappl mentioned this pull request Oct 28, 2024

Revert "[rtsan] Intercept aligned_alloc on all versions of OSX if available on build machine (#112780)" #113982

Merged

cjappl added a commit to cjappl/llvm-project that referenced this pull request Oct 29, 2024

Reapply "[rtsan] Intercept aligned_alloc on all versions of OSX if av…

f96f53f

…ailable on build machine (llvm#112780)" (llvm#113982) This reverts commit 7c55426.

cjappl mentioned this pull request Oct 29, 2024

Reland "[rtsan] Intercept aligned_alloc on all versions of OSX if available on the build machine" #114153

Merged

NoumanAmir657 pushed a commit to NoumanAmir657/llvm-project that referenced this pull request Nov 4, 2024

[rtsan] Intercept aligned_alloc on all versions of OSX if available o…

3a58767

…n build machine (llvm#112780)

	// dispatch_async_and_wait() and friends were introduced in macOS 10.14.
	// Linking of these interceptors fails when using an older SDK.
	#if !SANITIZER_APPLE \|\| defined(__MAC_10_14)
	// macOS 10.14 is greater than our minimal deployment target. To ensure we
	// generate a weak reference so the TSan dylib continues to work on older
	// systems, we need to forward declare the intercepted functions as "weak
	// imports". Note that this file is multi-platform, so we cannot include the
	// actual header file (#include <dispatch/dispatch.h>).
	SANITIZER_WEAK_IMPORT void dispatch_async_and_wait(
	dispatch_queue_t queue, DISPATCH_NOESCAPE dispatch_block_t block);
	SANITIZER_WEAK_IMPORT void dispatch_async_and_wait_f(
	dispatch_queue_t queue, void *context, dispatch_function_t work);
	SANITIZER_WEAK_IMPORT void dispatch_barrier_async_and_wait(
	dispatch_queue_t queue, DISPATCH_NOESCAPE dispatch_block_t block);
	SANITIZER_WEAK_IMPORT void dispatch_barrier_async_and_wait_f(
	dispatch_queue_t queue, void *context, dispatch_function_t work);

	DISPATCH_INTERCEPT_SYNC_F(dispatch_async_and_wait_f, false)
	DISPATCH_INTERCEPT_SYNC_B(dispatch_async_and_wait, false)
	DISPATCH_INTERCEPT_SYNC_F(dispatch_barrier_async_and_wait_f, true)
	DISPATCH_INTERCEPT_SYNC_B(dispatch_barrier_async_and_wait, true)
	#endif

[rtsan] Intercept aligned_alloc on all versions of OSX if available on build machine #112780

[rtsan] Intercept aligned_alloc on all versions of OSX if available on build machine #112780

Uh oh!

Conversation

cjappl commented Oct 17, 2024

Uh oh!

llvmbot commented Oct 17, 2024

Uh oh!

cjappl commented Oct 17, 2024

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

cjappl commented Oct 24, 2024

Uh oh!

Uh oh!

wrotki commented Oct 28, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

cjappl commented Oct 28, 2024

Uh oh!

wrotki commented Oct 28, 2024

Uh oh!

cjappl commented Oct 28, 2024

Uh oh!

wrotki commented Oct 28, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

cjappl commented Oct 28, 2024

Uh oh!

cjappl commented Oct 29, 2024

Uh oh!

wrotki commented Oct 29, 2024

Uh oh!

cjappl commented Oct 29, 2024

Uh oh!

Uh oh!

wrotki commented Oct 28, 2024 •

edited

Loading

wrotki commented Oct 28, 2024 •

edited

Loading