Now run-javascriptcore-tests --ftl-jit tests pass if libllvmForJSC.so isn't available. It isn't good that JSC silently falls back to non FTL path, because we can easily think that FTL is working well and all tests pass. ( It is related to bug143372 . ) Additionally now we can run FTL tests if we built JSC with disabled FTL. It would be good to get error in this case too.
(In reply to comment #0) > Now run-javascriptcore-tests --ftl-jit tests pass if libllvmForJSC.so isn't > available. > It isn't good that JSC silently falls back to non FTL path, because we can > easily > think that FTL is working well and all tests pass. ( It is related to > bug143372 . ) I think JSC should crash with an error message if we execute it with run-javascriptcore-tests and fall back to non FTL path in production code. > Additionally now we can run FTL tests if we built JSC with disabled FTL. > It would be good to get error in this case too. I'm going to fix it later in a separated bug.
Created attachment 250072 [details] Patch
Comment on attachment 250072 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=250072&action=review > Source/JavaScriptCore/dfg/DFGPlan.cpp:418 > + if (Options::ftlCrashesIfCantInitilizeLLVM()) { typo: initi_a_lize
Created attachment 250073 [details] Patch typo fixed
Comment on attachment 250073 [details] Patch View in context: https://bugs.webkit.org/attachment.cgi?id=250073&action=review > Source/JavaScriptCore/dfg/DFGPlan.cpp:418 > + if (Options::ftlCrashesIfCantInitilizeLLVM()) { typo: Initilize > Source/JavaScriptCore/runtime/Options.h:174 > + v(bool, ftlCrashesIfCantInitailizeLLVM, false, nullptr) \ typo: Initailize
Created attachment 250075 [details] Patch really fix the typo now
Comment on attachment 250075 [details] Patch r=me
Comment on attachment 250075 [details] Patch Clearing flags on attachment: 250075 Committed r182318: <http://trac.webkit.org/changeset/182318>
All reviewed patches have been landed. Closing bug.