70041 2011-10-13 12:10:27 -0700 IndexedDB: Transactions fail to abort if all requests throw exceptions 2011-11-16 11:37:58 -0800 1 1 1 Unclassified WebKit New Bugs 528+ (Nightly build) Unspecified Unspecified RESOLVED WORKSFORME P2 Normal --- 1 jsbell jsbell dgrogan gregsimon hans jsbell oldest_to_newest 483366 0 jsbell 2011-10-13 12:10:27 -0700 Repro: * Start a transaction * Make a request against the transaction that raises an exception (e.g. a put() with an invalid key) * Make no other requests against the transaction, let it execute Expected: * Transaction fires abort event Actual: * Transaction never fires abort, error or complete events Initial investigation: This appears to be caused by the path taken when requests are created but aborted early due to exceptions - a request is created, registered with the transaction, then marked for "early death". But the markEarlyDeath() method does not unregister it from the transaction; that doesn't happen until the request's destructor is called. Perhaps it should be unregistered immediately? Further investigation is needed, and we should understand why the transaction never completes. 483372 1 110885 jsbell 2011-10-13 12:13:21 -0700 Created attachment 110885 Repro case for the bug Attachment (bug70041.html) should show ABORTED (or ERROR or COMPLETE) but never does. You can try commenting out both store.put() to see it output ABORTED, or uncomment both store.put() calls to see it COMPLETE 483392 2 jsbell 2011-10-13 12:32:21 -0700 NOTE: LayoutTests/storage/indexeddb/objectstore-basics.html has an unexpected dependency on this buggy behavior - it creates a transaction that has no requests other than one which generated an exception, yet expects the transaction to not abort. So fixing this issue may cause that layout test to (erroneously) fail. 495856 3 jsbell 2011-11-03 16:23:58 -0700 The attached repro case no longer repros :( It reports "ABORTED" 499187 4 jsbell 2011-11-09 11:29:40 -0800 (In reply to comment #3) > The attached repro case no longer repros :( It reports "ABORTED" To clarify, it repros in M16, but not TOT Chrome. 110885 2011-10-13 12:13:21 -0700 2011-10-13 12:13:21 -0700 Repro case for the bug bug70041.html text/html 1073 jsbell PCFET0NUWVBFIGh0bWw+CjxzY3JpcHQ+CndpbmRvdy5pbmRleGVkREIgPSB3aW5kb3cuaW5kZXhl ZERCIHx8IHdpbmRvdy53ZWJraXRJbmRleGVkREI7CndpbmRvdy5JREJUcmFuc2FjdGlvbiA9IHdp bmRvdy5JREJUcmFuc2FjdGlvbiB8fCB3aW5kb3cud2Via2l0SURCVHJhbnNhY3Rpb247Cgp2YXIg b3BlbnJlcSA9IHdlYmtpdEluZGV4ZWREQi5vcGVuKCJkYiIgKyBTdHJpbmcobmV3IERhdGUpKTsK b3BlbnJlcS5vbmVycm9yID0gZnVuY3Rpb24oZSkgeyBjb25zb2xlLmVycm9yKGUpOyB9OwpvcGVu cmVxLm9uc3VjY2VzcyA9IGZ1bmN0aW9uKGUpIHsKICB2YXIgZGIgPSBlLnRhcmdldC5yZXN1bHQ7 CiAgaWYgKGRiLnZlcnNpb24gIT09ICIxIikgewogICAgdmFyIHN2cmVxID0gZGIuc2V0VmVyc2lv bigiMSIpOwogICAgc3ZyZXEub25lcnJvciA9IGZ1bmN0aW9uKGUpIHsgY29uc29sZS5lcnJvcihl KTsgfTsKICAgIHN2cmVxLm9uc3VjY2VzcyA9IGZ1bmN0aW9uIChlKSB7IAogICAgICBkYi5jcmVh dGVPYmplY3RTdG9yZSgic3RvcmUiKTsKICAgICAgZG9UcmFucygpOwogICAgfTsKICB9IGVsc2Ug ewogICAgZG9UcmFucygpOwogIH0KCiAgZnVuY3Rpb24gZG9UcmFucygpIHsKICAgIHZhciB0cmFu cyA9IGRiLnRyYW5zYWN0aW9uKFsic3RvcmUiXSwgSURCVHJhbnNhY3Rpb24uUkVBRF9XUklURSk7 CiAgICB2YXIgc3RvcmUgPSB0cmFucy5vYmplY3RTdG9yZSgnc3RvcmUnKTsKICAgIHRyeSB7IAog ICAgICAgIHN0b3JlLnB1dCgidmFsdWUiLCBudWxsKTsgLy8gaW52YWxpZCBrZXkKICAgIH0gY2F0 Y2ggKGUpIHsKICAgIH0KICAgIC8vc3RvcmUucHV0KCJhbm90aGVyIHZhbHVlIiwgInZhbGlkIGtl eSIpOwoKICAgIHRyYW5zLm9uYWJvcnQgPSBmdW5jdGlvbihlKSB7IGRvY3VtZW50LndyaXRlKCJB Qk9SVEVEIik7IH07CiAgICB0cmFucy5vbmVycm9yID0gZnVuY3Rpb24oZSkgeyBkb2N1bWVudC53 cml0ZSgiRVJST1IiKTsgfTsKICAgIHRyYW5zLm9uY29tcGxldGUgPSBmdW5jdGlvbihlKSB7IGRv Y3VtZW50LndyaXRlKCJDT01QTEVURSIpOyB9OwogIH0KCn07Cjwvc2NyaXB0Pgo=