Michael has landed: https://code.google.com/p/v8/source/detail?r=16623
This should unblock this patch (I've verified locally that GC stress no longer crashes). Adam, can you please attempt to re-land? On Thu, Sep 5, 2013 at 12:35 PM, Rafael Weinstein <[email protected]>wrote: > Note that I am now only seeing the crash under GC stress: > > /Volumes/src/v8/out/ia32.debug/d8 --test --nocrankshaft --nobreak-on-abort > --nodead-code-elimination --nofold-constants --enable-slow-asserts > --debug-code --verify-heap --harmony-observation --harmony-proxies > --harmony-collections --harmony-symbols --allow-natives-syntax > /Volumes/src/v8/test/mjsunit/mjsunit.js > /Volumes/src/v8/test/mjsunit/harmony/object-observe.js --gc-interval=500 > --stress-compaction --concurrent-recompilation-queue-length=64 > --concurrent-recompilation-delay=500 --concurrent-recompilation > > But it appears to be exactly the same stack trace. > > > On Wed, Sep 4, 2013 at 6:18 AM, Rafael Weinstein <[email protected]>wrote: > >> Any progress on this? >> >> >> On Thu, Aug 29, 2013 at 11:09 AM, Rafael Weinstein <[email protected]>wrote: >> >>> Thanks much! >>> >>> >>> On Thu, Aug 29, 2013 at 11:07 AM, Michael Starzinger < >>> [email protected]> wrote: >>> >>>> Hey Rafael! >>>> >>>> I already have a strong suspicion what is causing this. It is another >>>> instance of where we are calling from unhandlified into handlified code, >>>> but keep using stale handles. Toon and I are handlifying everything around >>>> JSObject::SetProperty right now, because it keeps biting us too often by >>>> now. I can reproduce the issue you mentioned in your mail and will probably >>>> have a fix for you tomorrow. >>>> >>>> Best regards, >>>> Michael >>>> >>>> >>>> On Thu, Aug 29, 2013 at 7:19 PM, Rafael Weinstein >>>> <[email protected]>wrote: >>>> >>>>> It looks like maybe in setting the value in the WeakMap, a hash is >>>>> created for the key value, but when it gets set, the resulting object's >>>>> map >>>>> is bad memory. >>>>> >>>>> Michael, any chance you can look at this? >>>>> >>>>> >>>>> On Wed, Aug 28, 2013 at 3:21 PM, Rafael Weinstein >>>>> <[email protected]>wrote: >>>>> >>>>>> So I've been able to narrow this down and make it more reproducible, >>>>>> but I'm out of my depth trying to figure out what's going on. It's >>>>>> crashing >>>>>> while trying to set a value in WeakMap (the map appears to be trashed). >>>>>> >>>>>> -Apply the reverted patch: >>>>>> https://code.google.com/p/v8/source/detail?r=16343 >>>>>> -Use a reduced object-observe.js test (attached) >>>>>> -Compile x64.debug >>>>>> -Run with standard test flags: --test --nocrankshaft >>>>>> --nobreak-on-abort --nodead-code-elimination --nofold-constants >>>>>> --enable-slow-asserts --debug-code --verify-heap --harmony-observation >>>>>> --harmony-proxies --harmony-collections --harmony-symbols >>>>>> --allow-natives-syntax >>>>>> >>>>>> I get: >>>>>> >>>>>> Program received signal EXC_BAD_ACCESS, Could not access memory. >>>>>> Reason: 13 at address: 0x0000000000000000 >>>>>> >>>>>> here: >>>>>> >>>>>> #0 v8::internal::Map::instance_type (this=0x1beefdad0beefdaf) at >>>>>> objects-inl.h:3467 >>>>>> #1 0x0000000100013984 in v8::internal::Object::IsOddball >>>>>> (this=0x21bd78effb21) at objects-inl.h:674 >>>>>> #2 0x0000000100043729 in v8::internal::Object::IsUndefined >>>>>> (this=0x21bd78effb21) at objects-inl.h:856 >>>>>> #3 0x00000001002edf7e in v8::internal::JSObject::GetIdentityHash >>>>>> (this=0x21bd78effb21, flag=v8::internal::ALLOW_CREATION) at >>>>>> ../src/objects.cc:4742 >>>>>> #4 0x000000010031acc6 in v8::internal::JSReceiver::GetIdentityHash >>>>>> (this=0x21bd78effb21, flag=v8::internal::ALLOW_CREATION) at >>>>>> objects-inl.h:5820 >>>>>> #5 0x00000001002da672 in v8::internal::Object::GetHash >>>>>> (this=0x21bd78effb21, flag=v8::internal::ALLOW_CREATION) at >>>>>> ../src/objects.cc:1018 >>>>>> #6 0x00000001002eecf3 in v8::internal::ObjectHashTable::Put >>>>>> (this=0x5a9e5c5b9f1, key=0x21bd78effb21, value=0x21bd78efff91) at >>>>>> ../src/objects.cc:15402 >>>>>> #7 0x000000010015cfdf in v8::internal::PutIntoObjectHashTable >>>>>> (table={location_ = 0x1018575f8}, key={location_ = 0x7fff5fbfee98}, >>>>>> value={location_ = 0x1018575f0}) at ../src/handles.cc:862 >>>>>> #8 0x00000001003834e9 in __RT_impl_Runtime_WeakCollectionSet >>>>>> (args={<v8::internal::Embedded> = {<No data fields>}, length_ = 3, >>>>>> arguments_ = 0x7fff5fbfeea0}, isolate=0x101803200) at >>>>>> ../src/runtime.cc:1565 >>>>>> #9 0x0000000100383386 in v8::internal::Runtime_WeakCollectionSet >>>>>> (args_length=3, args_object=0x7fff5fbfeea0, isolate=0x101803200) at >>>>>> ../src/runtime.cc:1557 >>>>>> #10 0x00000975806072ce in ?? () >>>>>> #11 0x0000097580649962 in ?? () >>>>>> #12 0x00000975806496a2 in ?? () >>>>>> #13 0x0000097580648ed9 in ?? () >>>>>> #14 0x0000097580610634 in ?? () >>>>>> #15 0x00000975806456f3 in ?? () >>>>>> #16 0x0000097580642764 in ?? () >>>>>> #17 0x000009758062e924 in ?? () >>>>>> #18 0x0000097580619817 in ?? () >>>>>> #19 0x00000001001072c9 in Invoke (is_construct=false, >>>>>> function={location_ = 0x1018575d0}, receiver={location_ = 0x1018575e0}, >>>>>> argc=0, args=0x0, has_pending_exception=0x7fff5fbff337) at >>>>>> ../src/execution.cc:119 >>>>>> #20 0x0000000100106d9a in v8::internal::Execution::Call >>>>>> (callable={location_ = 0x1018575d0}, receiver={location_ = 0x1018575e0}, >>>>>> argc=0, argv=0x0, pending_exception=0x7fff5fbff337, >>>>>> convert_receiver=false) >>>>>> at ../src/execution.cc:182 >>>>>> #21 0x0000000100025831 in v8::Script::Run (this=0x1018575a8) at >>>>>> ../src/api.cc:1848 >>>>>> #22 0x0000000100000f6e in v8::Shell::ExecuteString >>>>>> (isolate=0x101803200, source={val_ = 0x1018575a0}, name={val_ = >>>>>> 0x101857598}, print_result=false, report_exceptions=true) at >>>>>> ../src/d8.cc:216 >>>>>> #23 0x0000000100005ca4 in v8::SourceGroup::Execute (this=0x101003d78, >>>>>> isolate=0x101803200) at ../src/d8.cc:1257 >>>>>> #24 0x000000010000695e in v8::Shell::RunMain (isolate=0x101803200, >>>>>> argc=16, argv=0x7fff5fbff8f0) at ../src/d8.cc:1517 >>>>>> #25 0x0000000100006ded in v8::Shell::Main (argc=16, >>>>>> argv=0x7fff5fbff8f0) at ../src/d8.cc:1698 >>>>>> #26 0x000000010000be32 in main (argc=16, argv=0x7fff5fbff8f0) at >>>>>> ../src/d8.cc:173 >>>>>> >>>>>> >>>>>> On Tue, Aug 27, 2013 at 2:33 AM, <[email protected]> wrote: >>>>>> >>>>>>> On 2013/08/26 23:11:33, adamk - vacation until sept 3 wrote: >>>>>>> >>>>>>>> Tried runnings tests with >>>>>>>> >>>>>>> >>>>>>> "--extra-flags --noenable-sse2" >>>>>>>> >>>>>>> >>>>>>> which is what the bot seems to do, but the test still passes for me. >>>>>>>> >>>>>>> >>>>>>> Running this in the V8 dir seems to reproduce it (thanks to Michael): >>>>>>> >>>>>>> ./tools/run-tests.py --arch=ia32 --mode=release >>>>>>> --extra-flags="--noenable-**sse2" >>>>>>> mjsunit/harmony/object-observe >>>>>>> >>>>>>> However, it is flaky. You probably have to run it a few times before >>>>>>> the failure >>>>>>> occurs. >>>>>>> >>>>>>> Unfortunately, I have no idea off-hand concerning the cause. >>>>>>> >>>>>>> >>>>>>> https://codereview.chromium.**org/23491002/<https://codereview.chromium.org/23491002/> >>>>>>> >>>>>> >>>>>> >>>>> >>>> >>> >> > -- -- v8-dev mailing list [email protected] http://groups.google.com/group/v8-dev --- You received this message because you are subscribed to the Google Groups "v8-dev" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/groups/opt_out.
