(function(x) { x = +x switch (y) { case -1: x = 0 } })() asserts js debug shell on m-c changeset acde07cb4e4d with --ion-eager --no-threads --ion-gvn=off at Assertion failure: iter->instruction()->block() == *block, at jit/IonAnalysis.cpp. Debug configure options: CC="clang -Qunused-arguments" CXX="clang++ -Qunused-arguments" AR=ar AUTOCONF=/usr/local/Cellar/autoconf213/2.13/bin/autoconf213 sh /Users/skywalker/trees/mozilla-central/js/src/configure --target=x86_64-apple-darwin12.5.0 --enable-debug --enable-optimize --enable-nspr-build --enable-more-deterministic --with-ccache --enable-gczeal --enable-debug-symbols --disable-tests autoBisect shows this is probably related to the following changeset: The first bad revision is: changeset: https://hg.mozilla.org/mozilla-central/rev/9188c8b7962b user: Nicolas B. Pierron date: Mon Nov 24 16:11:32 2014 +0100 summary: Bug 1093674 - IonMonkey: Add Sink for instruction which can be recovered on bailout. r=sunfish Nicolas, is bug 1093674 a likely regressor?

(lldb) bt 5 * thread #1: tid = 0x48cbaa, 0x00000001002b625f js-dbg-opt-64-dm-nsprBuild-darwin-acde07cb4e4d`js::jit::AssertBasicGraphCoherency(js::jit::MIRGraph&) [inlined] js::InlineForwardListIterator<js::jit::MResumePoint>::operator*() const + 13 at InlineList.h:179, queue = 'com.apple.main-thread', stop reason = EXC_BAD_ACCESS (code=1, address=0x0) * frame #0: 0x00000001002b625f js-dbg-opt-64-dm-nsprBuild-darwin-acde07cb4e4d`js::jit::AssertBasicGraphCoherency(js::jit::MIRGraph&) [inlined] js::InlineForwardListIterator<js::jit::MResumePoint>::operator*() const + 13 at InlineList.h:179 frame #1: 0x00000001002b6252 js-dbg-opt-64-dm-nsprBuild-darwin-acde07cb4e4d`js::jit::AssertBasicGraphCoherency(graph=<unavailable>) + 3026 at IonAnalysis.cpp:1894 frame #2: 0x00000001002b744a js-dbg-opt-64-dm-nsprBuild-darwin-acde07cb4e4d`js::jit::AssertGraphCoherency(graph=0x00000001020bb040) + 42 at IonAnalysis.cpp:2033 frame #3: 0x00000001002b82dd js-dbg-opt-64-dm-nsprBuild-darwin-acde07cb4e4d`js::jit::AssertExtendedGraphCoherency(graph=0x00000001020bb040) + 45 at IonAnalysis.cpp:2063 frame #4: 0x00000001002b4f90 js-dbg-opt-64-dm-nsprBuild-darwin-acde07cb4e4d`js::jit::OptimizeMIR(mir=0x00000001020bb1a8) + 2752 at Ion.cpp:1571 (lldb)

(In reply to Gary Kwong [:gkw] [:nth10sd] from comment #0) > Nicolas, is bug 1093674 a likely regressor? I think so, I will look at it later today.

If the instruction is no longer effectful (such as the MMul of 1 after the apply type phase), then we can discard the resume point.

(try) https://treeherder.mozilla.org/ui/#/jobs?repo=try&revision=dcf94a9e3918 (inbound) https://hg.mozilla.org/integration/mozilla-inbound/rev/f41f1f33b1b3