141161 2015-02-02 08:23:27 -0800 Remove slow test unused code from run-jsc-stress-tests 2015-02-09 06:09:22 -0800 1 1 1 Unclassified WebKit New Bugs 528+ (Nightly build) Unspecified Unspecified RESOLVED INVALID P2 Normal --- 1 ossy ossy fpizlo ggaren mark.lam mhahnenb msaboff oliver ossy oldest_to_newest 1066100 0 ossy 2015-02-02 08:23:27 -0800 Handling slow tests mechanism was added by http://trac.webkit.org/changeset/157887 16 months before to prepend the only one slow test to the run list instead of append. LayoutTests/js/regress/script-tests/stepanov_container.js was removed from the trunk 13 months before - http://trac.webkit.org/changeset/162181 This hacky slow test mechanism wasn't used at all after it. I think we don't need it at all, let's clean up the code. 1066101 1 245878 ossy 2015-02-02 08:26:05 -0800 Created attachment 245878 Patch 1066102 2 245878 ossy 2015-02-02 08:26:33 -0800 Comment on attachment 245878 Patch wrong patch 1066103 3 245879 ossy 2015-02-02 08:27:30 -0800 Created attachment 245879 Patch 1067646 4 ossy 2015-02-07 23:05:34 -0800 ping? 1067666 5 245879 fpizlo 2015-02-08 06:51:24 -0800 Comment on attachment 245879 Patch Please don't remove this. :-). We use the --extra-tests flag to pass tests that cannot be made public for whatever reasons, and most of those use the slow mechanism. Plus I'm surprised that you think this is a hack. Of course you want slow-running tests to get pretended rather than appended, to aid with load balancing. 1067780 6 ossy 2015-02-09 02:41:22 -0800 (In reply to comment #5) > Comment on attachment 245879 [details] > Patch > > Please don't remove this. :-). We use the --extra-tests flag to pass tests > that cannot be made public for whatever reasons, and most of those use the > slow mechanism. OK, now I know it is used, I don't want to remove a used feature. But as I mentioned in the description, r157887 added it for only one test which was removed by r162181 and haven't found any reference if it is used. If a feature has only non-public use cases, it would be great to have some comment why is it needed in the code or in the bug report which added it. > Plus I'm surprised that you think this is a hack. Of course you want > slow-running tests to get pretended rather than appended, to aid with load > balancing. I still think and consider this code a hack. I understad the goal, but I'm surprised if it worked ever. Let's see the GNU make based runner: all: test_done_0 test_done_1 test_done_2 test_done_3 ... ... The order of the prerequisite doesn't guaranatee anything. If you want to make sure if test_done_0 runs before test_done_1, you have to add normal or order only prerequisite. Let's see the shell based runner. It collects the tests with external find call without any ordering and runner threads always pick up a test from the end of the list to run. I checked this list file, it is absolutely random on Linux and it is more or less random on OSX too: ./test_script_0 ./test_script_1 ./test_script_10 ./test_script_100 ./test_script_1000 ./test_script_10000 ./test_script_10001 ... ./test_script_10009 ./test_script_1001 ... ./test_script_16998 ./test_script_16999 ./test_script_17 ./test_script_170 ./test_script_1700 ./test_script_17000 ... ./test_script_17025 ./test_script_17026 ./test_script_1703 ./test_script_1704 ... ./test_script_1709 ./test_script_171 ... It seems as if it was sorted with sort, but not with numberic sort at all. 1067781 7 ossy 2015-02-09 02:42:34 -0800 Otherwise close as resolved/invalid, because it is still used. 1067785 8 ossy 2015-02-09 03:09:28 -0800 I fixed the shell test runner for you in bug141383. The make runner runs the tests in good order now on Linux and OSX too, but it isn't guaranteed at all, it is only a luck. 1067811 9 fpizlo 2015-02-09 06:09:22 -0800 (In reply to comment #8) > I fixed the shell test runner for you in bug141383. > > The make runner runs the tests in good order now > on Linux and OSX too, but it isn't guaranteed at > all, it is only a luck. Nope, not luck. GNU make has a predictable search that drives execution order. It's done it the same way for at least 15 years. I haven't seen it documented but they probably can't change it because it would break makefiles that accidentally depend on it. We don't care about the parallel test runner achieving great scalability on other make implementations. Also, all we need here is luck. Any test can be run in any order, but we scale better if the slowest tests are fired off first. So if the make we ran on decided to break with tradition and use a different ordering then we would just lose the efficacy of the slow test optimization but this wouldn't affect the validity of the test run. 245878 2015-02-02 08:26:05 -0800 2015-02-02 08:26:33 -0800 Patch bug-141161-20150202082613.patch text/plain 3501 ossy U3VidmVyc2lvbiBSZXZpc2lvbjogMTc5NDc0CmRpZmYgLS1naXQgYS9Ub29scy9DaGFuZ2VMb2cg Yi9Ub29scy9DaGFuZ2VMb2cKaW5kZXggNmU1ZmUyNDk1YTU4YzM4MTI2MDI3NWQ3MDU1YzA5Mjhj YTllMDM3ZC4uOGFkNDQzZmJiYjA1ZTA4N2EyYTRjMTg3NmMxZTRhMTMzNjZlMjU5NyAxMDA2NDQK LS0tIGEvVG9vbHMvQ2hhbmdlTG9nCisrKyBiL1Rvb2xzL0NoYW5nZUxvZwpAQCAtMSwzICsxLDEy IEBACisyMDE1LTAyLTAyICBDc2FiYSBPc3p0cm9nb27DoWMgIDxvc3N5QHdlYmtpdC5vcmc+CisK KyAgICAgICAgUmVtb3ZlIHNsb3cgdGVzdCB1bnVzZWQgY29kZSBmcm9tIHJ1bi1qc2Mtc3RyZXNz LXRlc3RzCisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0x NDExNjEKKworICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICAq IFNjcmlwdHMvcnVuLWpzYy1zdHJlc3MtdGVzdHM6CisKIDIwMTUtMDItMDIgIE1pY2hhZWwgQ2F0 YW56YXJvICA8bWNhdGFuemFyb0BpZ2FsaWEuY29tPgogCiAgICAgICAgIFtHVEtdIE1pbmlCcm93 c2VyIHNob3VsZCBjbG9zZSBpdHNlbGYgb24gQ3RybCtXIG9yIEN0cmwrUQpkaWZmIC0tZ2l0IGEv VG9vbHMvU2NyaXB0cy9ydW4tanNjLXN0cmVzcy10ZXN0cyBiL1Rvb2xzL1NjcmlwdHMvcnVuLWpz Yy1zdHJlc3MtdGVzdHMKaW5kZXggNjA3YzgzYzE1ZWQ0ZjhkYTg4MGVjZTg0MWU2NTM2NzBlNDg5 ZWExNi4uMTYyMzQ5NTllODM1ZjdlNzY5NGYxOWExZGRkYTQ0OTg2YjIxOTExNiAxMDA3NTUKLS0t IGEvVG9vbHMvU2NyaXB0cy9ydW4tanNjLXN0cmVzcy10ZXN0cworKysgYi9Ub29scy9TY3JpcHRz L3J1bi1qc2Mtc3RyZXNzLXRlc3RzCkBAIC00NTEsOCArNDUxLDYgQEAgZGVmIG1vemlsbGFFeGl0 M0Vycm9ySGFuZGxlcgogICAgIH0KIGVuZAogCi0kcnVuQ29tbWFuZE9wdGlvbnMgPSB7fQotCiBj bGFzcyBQbGFuCiAgICAgYXR0cl9yZWFkZXIgOmRpcmVjdG9yeSwgOmFyZ3VtZW50cywgOm5hbWUs IDpvdXRwdXRIYW5kbGVyLCA6ZXJyb3JIYW5kbGVyCiAgICAgYXR0cl9hY2Nlc3NvciA6aW5kZXgK QEAgLTQ2Myw3ICs0NjEsNiBAQCBjbGFzcyBQbGFuCiAgICAgICAgIEBuYW1lID0gbmFtZQogICAg ICAgICBAb3V0cHV0SGFuZGxlciA9IG91dHB1dEhhbmRsZXIKICAgICAgICAgQGVycm9ySGFuZGxl ciA9IGVycm9ySGFuZGxlcgotICAgICAgICBAaXNTbG93ID0gISEkcnVuQ29tbWFuZE9wdGlvbnNb OmlzU2xvd10KICAgICBlbmQKICAgICAKICAgICBkZWYgc2hlbGxDb21tYW5kCkBAIC01NDEsMTIg KzUzOCw3IEBAIGVuZAogCiBkZWYgYWRkUnVuQ29tbWFuZChraW5kLCBjb21tYW5kLCBvdXRwdXRI YW5kbGVyLCBlcnJvckhhbmRsZXIpCiAgICAgJGRpZEFkZFJ1bkNvbW1hbmQgPSB0cnVlCi0gICAg cGxhbiA9IFBsYW4ubmV3KCRiZW5jaG1hcmtEaXJlY3RvcnksIGNvbW1hbmQsIGJhc2VPdXRwdXRO YW1lKGtpbmQpLCBvdXRwdXRIYW5kbGVyLCBlcnJvckhhbmRsZXIpCi0gICAgaWYgJG51bVByb2Nl c3NvcnMgPiAxIGFuZCAkcnVuQ29tbWFuZE9wdGlvbnNbOmlzU2xvd10KLSAgICAgICAgJHJ1bmxp c3QudW5zaGlmdCBwbGFuCi0gICAgZWxzZQotICAgICAgICAkcnVubGlzdCA8PCBwbGFuCi0gICAg ZW5kCisgICAgJHJ1bmxpc3QgPDwgUGxhbi5uZXcoJGJlbmNobWFya0RpcmVjdG9yeSwgY29tbWFu ZCwgYmFzZU91dHB1dE5hbWUoa2luZCksIG91dHB1dEhhbmRsZXIsIGVycm9ySGFuZGxlcikKIGVu ZAogCiAjIFJldHVybnMgdHJ1ZSBpZiB0aGVyZSB3ZXJlIHJ1biBjb21tYW5kcyBmb3VuZCBpbiB0 aGUgZmlsZSAoJGJlbmNobWFya0RpcmVjdG9yeSArCkBAIC01NzksMTAgKzU3MSw2IEBAIGRlZiBw YXJzZVJ1bkNvbW1hbmRzCiAgICAgcmVzdWx0CiBlbmQKIAotZGVmIHNsb3chCi0gICAgJHJ1bkNv bW1hbmRPcHRpb25zWzppc1Nsb3ddID0gdHJ1ZQotZW5kCi0KIGRlZiBydW4oa2luZCwgKm9wdGlv bnMpCiAgICAgYWRkUnVuQ29tbWFuZChraW5kLCBbcGF0aFRvVk0udG9fc10gKyBCQVNFX09QVElP TlMgKyBvcHRpb25zICsgWyRiZW5jaG1hcmsudG9fc10sIHNpbGVudE91dHB1dEhhbmRsZXIsIHNp bXBsZUVycm9ySGFuZGxlcikKIGVuZApAQCAtMTAwNSw3ICs5OTMsNiBAQCBkZWYgaGFuZGxlQ29s bGVjdGlvbkZpbGUoY29sbGVjdGlvbikKICAgICAgICAgICAgICAgICAkYmVuY2htYXJrID0gcGF0 aC5iYXNlbmFtZQogICAgICAgICAgICAgICAgICRiZW5jaG1hcmtEaXJlY3RvcnkgPSBwYXRoLmRp cm5hbWUKICAgICAgICAgICAgICAgICAKLSAgICAgICAgICAgICAgICAkcnVuQ29tbWFuZE9wdGlv bnMgPSB7fQogICAgICAgICAgICAgICAgIGV2YWwgZW50cnlbImNtZCJdCiAgICAgICAgICAgICB9 CiAgICAgICAgIH0KQEAgLTEwMzEsNyArMTAxOCw2IEBAIGRlZiBoYW5kbGVDb2xsZWN0aW9uRGly ZWN0b3J5KGNvbGxlY3Rpb24pCiAgICAgICAgICAgICAKICAgICAgICAgICAgICRiZW5jaG1hcmsg PSBwYXRoLmJhc2VuYW1lCiAgICAgICAgICAgICAKLSAgICAgICAgICAgICRydW5Db21tYW5kT3B0 aW9ucyA9IHt9CiAgICAgICAgICAgICBkZWZhdWx0UnVuIHVubGVzcyBwYXJzZVJ1bkNvbW1hbmRz CiAgICAgICAgIH0KICAgICB9CkBAIC0xMzY3LDcgKzEzNTMsMTQgQEAgZGVmIHJ1blNoZWxsVGVz dFJ1bm5lcgogICAgICAgICByZW1vdGVTY3JpcHQgKz0gInJtIC1yZiAjeyRvdXRwdXREaXIuYmFz ZW5hbWV9ICYmICIKICAgICAgICAgcmVtb3RlU2NyaXB0ICs9ICJ0YXIgeHpmICN7JHRhckZpbGVO YW1lfSAmJiAiCiAgICAgICAgIHJlbW90ZVNjcmlwdCArPSAiY2QgI3skb3V0cHV0RGlyLmJhc2Vu YW1lfS8ucnVubmVyICYmICIKLSAgICAgICAgcmVtb3RlU2NyaXB0ICs9ICJEWUxEX0ZSQU1FV09S S19QQVRIPVxcXCJcXCQoY2QgI3skdGVzdGluZ0ZyYW1ld29ya1BhdGguZGlybmFtZX07IHB3ZClc XFwiIHNoIHJ1bnNjcmlwdCIKKyAgICAgICAgSU1QT1JUQU5UX0VOVlMuZWFjaCB7CisgICAgICAg ICAgICB8IGtleSB8CisgICAgICAgICAgICBpZiBFTlZba2V5XQorICAgICAgICAgICAgICAgIHJl bW90ZVNjcmlwdCArPSAiZXhwb3J0ICN7a2V5fT0je1NoZWxsd29yZHMuc2hlbGxlc2NhcGUoRU5W W2tleV0pfSAmJiAiCisgICAgICAgICAgICBlbmQKKyAgICAgICAgfQorICAgICAgICByZW1vdGVT Y3JpcHQgKz0gImV4cG9ydCBEWUxEX0ZSQU1FV09SS19QQVRIPVxcXCJcXCQoY2QgI3skdGVzdGlu Z0ZyYW1ld29ya1BhdGguZGlybmFtZX07IHB3ZClcXFwiICYmICIKKyAgICAgICAgcmVtb3RlU2Ny aXB0ICs9ICJzaCBydW5zY3JpcHQiCiAgICAgICAgIHJlbW90ZVNjcmlwdCArPSAiXCIiCiAgICAg ICAgIHJ1bkFuZE1vbml0b3JUZXN0UnVubmVyQ29tbWFuZChudW1iZXJPZlRlc3RzLCAic3NoIiwg Ii1wIiwgJHJlbW90ZVBvcnQudG9fcywgIiN7JHJlbW90ZVVzZXJ9QCN7JHJlbW90ZUhvc3R9Iiwg cmVtb3RlU2NyaXB0KQogICAgIGVsc2UK 245879 2015-02-02 08:27:30 -0800 2015-02-08 06:51:24 -0800 Patch bug-141161-20150202082737.patch text/plain 2641 ossy U3VidmVyc2lvbiBSZXZpc2lvbjogMTc5NDc0CmRpZmYgLS1naXQgYS9Ub29scy9DaGFuZ2VMb2cg Yi9Ub29scy9DaGFuZ2VMb2cKaW5kZXggNmU1ZmUyNDk1YTU4YzM4MTI2MDI3NWQ3MDU1YzA5Mjhj YTllMDM3ZC4uOGFkNDQzZmJiYjA1ZTA4N2EyYTRjMTg3NmMxZTRhMTMzNjZlMjU5NyAxMDA2NDQK LS0tIGEvVG9vbHMvQ2hhbmdlTG9nCisrKyBiL1Rvb2xzL0NoYW5nZUxvZwpAQCAtMSwzICsxLDEy IEBACisyMDE1LTAyLTAyICBDc2FiYSBPc3p0cm9nb27DoWMgIDxvc3N5QHdlYmtpdC5vcmc+CisK KyAgICAgICAgUmVtb3ZlIHNsb3cgdGVzdCB1bnVzZWQgY29kZSBmcm9tIHJ1bi1qc2Mtc3RyZXNz LXRlc3RzCisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0x NDExNjEKKworICAgICAgICBSZXZpZXdlZCBieSBOT0JPRFkgKE9PUFMhKS4KKworICAgICAgICAq IFNjcmlwdHMvcnVuLWpzYy1zdHJlc3MtdGVzdHM6CisKIDIwMTUtMDItMDIgIE1pY2hhZWwgQ2F0 YW56YXJvICA8bWNhdGFuemFyb0BpZ2FsaWEuY29tPgogCiAgICAgICAgIFtHVEtdIE1pbmlCcm93 c2VyIHNob3VsZCBjbG9zZSBpdHNlbGYgb24gQ3RybCtXIG9yIEN0cmwrUQpkaWZmIC0tZ2l0IGEv VG9vbHMvU2NyaXB0cy9ydW4tanNjLXN0cmVzcy10ZXN0cyBiL1Rvb2xzL1NjcmlwdHMvcnVuLWpz Yy1zdHJlc3MtdGVzdHMKaW5kZXggNjA3YzgzYzE1ZWQ0ZjhkYTg4MGVjZTg0MWU2NTM2NzBlNDg5 ZWExNi4uYzRlNDg0YTIzM2FkOGRmZWI4MmM2OGYyZjQyYWU4YTBlOGM3MGVmYyAxMDA3NTUKLS0t IGEvVG9vbHMvU2NyaXB0cy9ydW4tanNjLXN0cmVzcy10ZXN0cworKysgYi9Ub29scy9TY3JpcHRz L3J1bi1qc2Mtc3RyZXNzLXRlc3RzCkBAIC00NTEsOCArNDUxLDYgQEAgZGVmIG1vemlsbGFFeGl0 M0Vycm9ySGFuZGxlcgogICAgIH0KIGVuZAogCi0kcnVuQ29tbWFuZE9wdGlvbnMgPSB7fQotCiBj bGFzcyBQbGFuCiAgICAgYXR0cl9yZWFkZXIgOmRpcmVjdG9yeSwgOmFyZ3VtZW50cywgOm5hbWUs IDpvdXRwdXRIYW5kbGVyLCA6ZXJyb3JIYW5kbGVyCiAgICAgYXR0cl9hY2Nlc3NvciA6aW5kZXgK QEAgLTQ2Myw3ICs0NjEsNiBAQCBjbGFzcyBQbGFuCiAgICAgICAgIEBuYW1lID0gbmFtZQogICAg ICAgICBAb3V0cHV0SGFuZGxlciA9IG91dHB1dEhhbmRsZXIKICAgICAgICAgQGVycm9ySGFuZGxl ciA9IGVycm9ySGFuZGxlcgotICAgICAgICBAaXNTbG93ID0gISEkcnVuQ29tbWFuZE9wdGlvbnNb OmlzU2xvd10KICAgICBlbmQKICAgICAKICAgICBkZWYgc2hlbGxDb21tYW5kCkBAIC01NDEsMTIg KzUzOCw3IEBAIGVuZAogCiBkZWYgYWRkUnVuQ29tbWFuZChraW5kLCBjb21tYW5kLCBvdXRwdXRI YW5kbGVyLCBlcnJvckhhbmRsZXIpCiAgICAgJGRpZEFkZFJ1bkNvbW1hbmQgPSB0cnVlCi0gICAg cGxhbiA9IFBsYW4ubmV3KCRiZW5jaG1hcmtEaXJlY3RvcnksIGNvbW1hbmQsIGJhc2VPdXRwdXRO YW1lKGtpbmQpLCBvdXRwdXRIYW5kbGVyLCBlcnJvckhhbmRsZXIpCi0gICAgaWYgJG51bVByb2Nl c3NvcnMgPiAxIGFuZCAkcnVuQ29tbWFuZE9wdGlvbnNbOmlzU2xvd10KLSAgICAgICAgJHJ1bmxp c3QudW5zaGlmdCBwbGFuCi0gICAgZWxzZQotICAgICAgICAkcnVubGlzdCA8PCBwbGFuCi0gICAg ZW5kCisgICAgJHJ1bmxpc3QgPDwgUGxhbi5uZXcoJGJlbmNobWFya0RpcmVjdG9yeSwgY29tbWFu ZCwgYmFzZU91dHB1dE5hbWUoa2luZCksIG91dHB1dEhhbmRsZXIsIGVycm9ySGFuZGxlcikKIGVu ZAogCiAjIFJldHVybnMgdHJ1ZSBpZiB0aGVyZSB3ZXJlIHJ1biBjb21tYW5kcyBmb3VuZCBpbiB0 aGUgZmlsZSAoJGJlbmNobWFya0RpcmVjdG9yeSArCkBAIC01NzksMTAgKzU3MSw2IEBAIGRlZiBw YXJzZVJ1bkNvbW1hbmRzCiAgICAgcmVzdWx0CiBlbmQKIAotZGVmIHNsb3chCi0gICAgJHJ1bkNv bW1hbmRPcHRpb25zWzppc1Nsb3ddID0gdHJ1ZQotZW5kCi0KIGRlZiBydW4oa2luZCwgKm9wdGlv bnMpCiAgICAgYWRkUnVuQ29tbWFuZChraW5kLCBbcGF0aFRvVk0udG9fc10gKyBCQVNFX09QVElP TlMgKyBvcHRpb25zICsgWyRiZW5jaG1hcmsudG9fc10sIHNpbGVudE91dHB1dEhhbmRsZXIsIHNp bXBsZUVycm9ySGFuZGxlcikKIGVuZApAQCAtMTAwNSw3ICs5OTMsNiBAQCBkZWYgaGFuZGxlQ29s bGVjdGlvbkZpbGUoY29sbGVjdGlvbikKICAgICAgICAgICAgICAgICAkYmVuY2htYXJrID0gcGF0 aC5iYXNlbmFtZQogICAgICAgICAgICAgICAgICRiZW5jaG1hcmtEaXJlY3RvcnkgPSBwYXRoLmRp cm5hbWUKICAgICAgICAgICAgICAgICAKLSAgICAgICAgICAgICAgICAkcnVuQ29tbWFuZE9wdGlv bnMgPSB7fQogICAgICAgICAgICAgICAgIGV2YWwgZW50cnlbImNtZCJdCiAgICAgICAgICAgICB9 CiAgICAgICAgIH0KQEAgLTEwMzEsNyArMTAxOCw2IEBAIGRlZiBoYW5kbGVDb2xsZWN0aW9uRGly ZWN0b3J5KGNvbGxlY3Rpb24pCiAgICAgICAgICAgICAKICAgICAgICAgICAgICRiZW5jaG1hcmsg PSBwYXRoLmJhc2VuYW1lCiAgICAgICAgICAgICAKLSAgICAgICAgICAgICRydW5Db21tYW5kT3B0 aW9ucyA9IHt9CiAgICAgICAgICAgICBkZWZhdWx0UnVuIHVubGVzcyBwYXJzZVJ1bkNvbW1hbmRz CiAgICAgICAgIH0KICAgICB9Cg==