36934 2010-03-31 23:53:33 -0700 REGRESSION(56394): http/tests/xmlhttprequest/xmlhttprequest-50ms-download-dispatch.html fails on Tiger 2010-04-05 10:30:36 -0700 1 1 1 Unclassified WebKit Tools / Tests 528+ (Nightly build) Mac OS X 10.4 RESOLVED FIXED P2 Normal --- 18654 1 eric webkit-unassigned abarth ap commit-queue jchaffraix oldest_to_newest 206996 0 eric 2010-03-31 23:53:33 -0700 REGRESSION(56394): http/tests/xmlhttprequest/xmlhttprequest-50ms-download-dispatch.html fails on Tiger This test has been failing since it was committed. It may just need Tiger-specific results checked in, but that is very unclear from the (not very helpful) pass/fail output. http://build.webkit.org/results/Tiger%20Intel%20Release/r56893%20(10281)/http/tests/xmlhttprequest/xmlhttprequest-50ms-download-dispatch-diffs.txt --- /Volumes/Data/WebKit-BuildSlave/tiger-intel-release/build/layout-test-results/http/tests/xmlhttprequest/xmlhttprequest-50ms-download-dispatch-expected.txt 2010-03-31 21:39:46.000000000 -0700 +++ /Volumes/Data/WebKit-BuildSlave/tiger-intel-release/build/layout-test-results/http/tests/xmlhttprequest/xmlhttprequest-50ms-download-dispatch-actual.txt 2010-03-31 21:39:46.000000000 -0700 @@ -5,8 +5,8 @@ You should see a sequence of 5 PASSED. PASSED +FAILED (count was 1) PASSED PASSED -PASSED -PASSED +FAILED (count was 1) 206997 1 eric 2010-03-31 23:54:30 -0700 I see, the diff is just hard to read. Tiger is simply failing 2 of the 5 subtests. My "not very helpful" comment above was itself not very helpful. What I meant was that the PASSED/FAILED messages were not self-documenting. 207169 2 jchaffraix 2010-04-01 08:37:36 -0700 (In reply to comment #1) > I see, the diff is just hard to read. Tiger is simply failing 2 of the 5 > subtests. My "not very helpful" comment above was itself not very helpful. > What I meant was that the PASSED/FAILED messages were not self-documenting. You are right about the message. I will attach a patch to have more explicit messages that will help find out what is failing. It is likely the 2 tests that fails are the 2 throttled ones, the other ones just make sure there is no throttling. Thanks for filing this bug. I will be more careful with the Tiger build-bot next time. 207172 3 52299 jchaffraix 2010-04-01 08:41:47 -0700 Created attachment 52299 Add more explicit comments to the test for debugging this issue 207267 4 52299 eric 2010-04-01 11:22:41 -0700 Comment on attachment 52299 Add more explicit comments to the test for debugging this issue We would need to update the expected results when checking this in if we didn't want failures. 207289 5 jchaffraix 2010-04-01 11:49:51 -0700 (In reply to comment #4) > (From update of attachment 52299 [details]) > We would need to update the expected results when checking this in if we didn't > want failures. No, I only made the failures more clear. The test does asynchronous XHRs which means that saying which one passed is not reliable 100% of the time. There is value in keeping it as such (like checking that the throttling will still work properly with several concurrent XHRs). That's why I did not want to update the expected results. 207294 6 52299 eric 2010-04-01 11:57:26 -0700 Comment on attachment 52299 Add more explicit comments to the test for debugging this issue OK. 207310 7 52299 commit-queue 2010-04-01 12:23:45 -0700 Comment on attachment 52299 Add more explicit comments to the test for debugging this issue Clearing flags on attachment: 52299 Committed r56928: <http://trac.webkit.org/changeset/56928> 207311 8 commit-queue 2010-04-01 12:23:52 -0700 All reviewed patches have been landed. Closing bug. 207609 9 jchaffraix 2010-04-01 20:15:38 -0700 Reopening this bug as it was not mean to be closed. The Tiger build bot fails consistently 2 strict tests (my previous analysis was completely off): FAILED (count was 1) for strict test 1 FAILED (count was 1) for strict test 3 Looking at http/tests/xmlhttprequest/xmlhttprequest-50ms-download-dispatch.html, the only strict test that passes is expecting 1 'progress' event. The throttled tests would also pass with only 1 event. Without more information, it seems like the network library buffers the input. Increasing the time between chunks of data could help but we may miss some corner cases at the 50 ms limit (which would be difficult to get reliably on the bots). I will go ahead and increase it in the following days to make the bot happy. 207613 10 eric 2010-04-01 20:20:46 -0700 I do seem to remember Tiger's CFNetwork having some sort of buffer for loading http requests off the network. But I could be remembering wrong. AP or some other Apple person would know. 207622 11 ap 2010-04-01 20:44:17 -0700 Yes, our code that disables content sniffing (and thus buffering) only works on Leopard and later. Good catch! 207774 12 eric 2010-04-02 00:36:44 -0700 (In reply to comment #11) > Yes, our code that disables content sniffing (and thus buffering) only works on > Leopard and later. Good catch! So I'm unclear from your remark. Is there a code fix to make here, or just a test fix? 207860 13 ap 2010-04-02 08:27:17 -0700 I was answering your question in comment 10. Certainly XMLHttpRequest behavior is different on Tiger because of buffering, but I haven't investigated if test failures are caused by this difference. 207896 14 eric 2010-04-02 09:58:14 -0700 Adam Barth fixed another failing test in Tiger last night by moving it from using GET to POST. http://trac.webkit.org/changeset/56979 I'm not sure if the same workaround is useful in this context or not. 208094 15 jchaffraix 2010-04-02 15:17:22 -0700 (In reply to comment #14) > Adam Barth fixed another failing test in Tiger last night by moving it from > using GET to POST. http://trac.webkit.org/changeset/56979 I'm not sure if the > same workaround is useful in this context or not. I think this would apply here too. I tried such a change and it did not make the test fail. I will create a patch tonight (I cannot come back to that before). 208098 16 ap 2010-04-02 15:20:36 -0700 Oh, I think that we could perhaps use a non-text MIME type to disable buffering. My recollection is that CFNetwork only does content sniffing for types like text/plain, because that often stands for "unknown". 208236 17 52483 abarth 2010-04-02 23:03:50 -0700 Created attachment 52483 Patch 208237 18 52483 abarth 2010-04-02 23:15:12 -0700 Comment on attachment 52483 Patch Clearing flags on attachment: 52483 Committed r57044: <http://trac.webkit.org/changeset/57044> 208240 19 abarth 2010-04-02 23:34:51 -0700 ap for the win. Seems to have worked. Welcome to a green Tiger bot. 208542 20 ap 2010-04-05 10:26:02 -0700 Great! Can this bug be marked as resolved now? 208548 21 eric 2010-04-05 10:30:36 -0700 Looks resolved to me. 52299 2010-04-01 08:41:47 -0700 2010-04-02 23:03:46 -0700 Add more explicit comments to the test for debugging this issue 36934-explicit-comments.diff text/plain 3238 jchaffraix ZGlmZiAtLWdpdCBhL0xheW91dFRlc3RzL0NoYW5nZUxvZyBiL0xheW91dFRlc3RzL0NoYW5nZUxv ZwppbmRleCBjMDBmMzQ0Li4zMDI2M2JlIDEwMDY0NAotLS0gYS9MYXlvdXRUZXN0cy9DaGFuZ2VM b2cKKysrIGIvTGF5b3V0VGVzdHMvQ2hhbmdlTG9nCkBAIC0xLDMgKzEsMTYgQEAKKzIwMTAtMDQt MDEgIEp1bGllbiBDaGFmZnJhaXggIDxqY2hhZmZyYWl4QHdlYmtpdC5vcmc+CisKKyAgICAgICAg UmV2aWV3ZWQgYnkgTk9CT0RZIChPT1BTISkuCisKKyAgICAgICAgUGFydCBvZjoKKyAgICAgICAg UkVHUkVTU0lPTig1NjM5NCk6CisgICAgICAgIGh0dHAvdGVzdHMveG1saHR0cHJlcXVlc3QveG1s aHR0cHJlcXVlc3QtNTBtcy1kb3dubG9hZC1kaXNwYXRjaC5odG1sCisgICAgICAgIGZhaWxzIG9u IFRpZ2VyCisgICAgICAgIGh0dHBzOi8vYnVncy53ZWJraXQub3JnL3Nob3dfYnVnLmNnaT9pZD0z NjkzNAorCisgICAgICAgICogaHR0cC90ZXN0cy94bWxodHRwcmVxdWVzdC94bWxodHRwcmVxdWVz dC01MG1zLWRvd25sb2FkLWRpc3BhdGNoLmh0bWw6CisgICAgICAgIEFkZCBtb3JlIGV4cGxpY2l0 IGZhaWxpbmcgbWVzc2FnZXMgdG8gaGVscCBmaW5kIG91dCB3aGF0IGlzIGZhaWxpbmcuCisKIDIw MTAtMDQtMDEgIHlhZWwgYWhhcm9uICA8eWFlbC5haGFyb25Abm9raWEuY29tPgogCiAgICAgICAg IFJldmlld2VkIGJ5IExhc3psbyBHb21ib3MuCmRpZmYgLS1naXQgYS9MYXlvdXRUZXN0cy9odHRw L3Rlc3RzL3htbGh0dHByZXF1ZXN0L3htbGh0dHByZXF1ZXN0LTUwbXMtZG93bmxvYWQtZGlzcGF0 Y2guaHRtbCBiL0xheW91dFRlc3RzL2h0dHAvdGVzdHMveG1saHR0cHJlcXVlc3QveG1saHR0cHJl cXVlc3QtNTBtcy1kb3dubG9hZC1kaXNwYXRjaC5odG1sCmluZGV4IGQ5ODIwYzMuLjk2NjQ2MGYg MTAwNjQ0Ci0tLSBhL0xheW91dFRlc3RzL2h0dHAvdGVzdHMveG1saHR0cHJlcXVlc3QveG1saHR0 cHJlcXVlc3QtNTBtcy1kb3dubG9hZC1kaXNwYXRjaC5odG1sCisrKyBiL0xheW91dFRlc3RzL2h0 dHAvdGVzdHMveG1saHR0cHJlcXVlc3QveG1saHR0cHJlcXVlc3QtNTBtcy1kb3dubG9hZC1kaXNw YXRjaC5odG1sCkBAIC0yMCwyMSArMjAsMjEgQEAgZnVuY3Rpb24gbG9nKG1lc3NhZ2UpCiAgICAg ZG9jdW1lbnQuZ2V0RWxlbWVudEJ5SWQoImNvbnNvbGUiKS5hcHBlbmRDaGlsZChkb2N1bWVudC5j cmVhdGVFbGVtZW50KCJiciIpKTsKIH0KIAotZnVuY3Rpb24gdGVzdChpdGVyYXRpb24sIGRlbGF5 LCBjb21wYXJlKQorZnVuY3Rpb24gdGVzdChpdGVyYXRpb24sIGRlbGF5LCBjb21wYXJlLCB0ZXN0 RGVzY3JpcHRpb24pCiB7CiAgICAgdmFyIGNvdW50ID0gMDsKICAgICB2YXIgc2F3UmVhZHlTdGF0 ZURPTkUgPSBmYWxzZTsKICAgICBmdW5jdGlvbiBvblByb2dyZXNzKGUpIHsKICAgICAgICAgKytj b3VudDsKICAgICAgICAgaWYgKHNhd1JlYWR5U3RhdGVET05FKQotICAgICAgICAgICAgbG9nKCJG QUlMRUQ6IHNhdyAncHJvZ3Jlc3MnIGV2ZW50IGFmdGVyIHJlYWR5c3RhdGUgJ0RPTkUnIGV2ZW50 LiIpOworICAgICAgICAgICAgbG9nKCJGQUlMRUQ6IHNhdyAncHJvZ3Jlc3MnIGV2ZW50IGFmdGVy IHJlYWR5c3RhdGUgJ0RPTkUnIGV2ZW50IGZvciAiICsgdGVzdERlc2NyaXB0aW9uKTsKICAgICB9 CiAKICAgICBmdW5jdGlvbiBvblJlYWR5U3RhdGUoZSkgewogICAgICAgICBpZiAodGhpcy5yZWFk eVN0YXRlID09IDQpIHsKICAgICAgICAgICAgIHNhd1JlYWR5U3RhdGVET05FID0gdHJ1ZTsKICAg ICAgICAgICAgIHZhciBwYXNzZWQgPSAgY29tcGFyZShjb3VudCwgaXRlcmF0aW9uKTsKLSAgICAg ICAgICAgIGxvZyhwYXNzZWQgPyAiUEFTU0VEIiA6ICJGQUlMRUQgKGNvdW50IHdhcyAiICsgY291 bnQgKyAiKSIpOworICAgICAgICAgICAgbG9nKHBhc3NlZCA/ICJQQVNTRUQiIDogIkZBSUxFRCAo Y291bnQgd2FzICIgKyBjb3VudCArICIpIGZvciAiICsgdGVzdERlc2NyaXB0aW9uKTsKICAgICAg ICAgfQogICAgIH0KIApAQCAtNDUsNiArNDUsNyBAQCBmdW5jdGlvbiB0ZXN0KGl0ZXJhdGlvbiwg ZGVsYXksIGNvbXBhcmUpCiAgICAgcmVxLnNlbmQobnVsbCk7CiB9CiAKK3RyeSB7CiAvLyBOdW1i ZXIgb2YgY2h1bmtzIHRvIHNlbmQsIGRlbGF5IGJldHdlZW4gY2h1bmtzCiB2YXIgc3RyaWN0VGVz dHMgPSBbIDIsIDgwLAogICAgICAgICAgICAgICAgICAgICAxLCAxMDAwLApAQCAtNTQsNiArNTUs MTQgQEAgZnVuY3Rpb24gY29tcGFyZVN0cmljdChjb3VudCwgaXRlcmF0aW9uKQogICAgIHJldHVy biBjb3VudCA9PSBpdGVyYXRpb247CiB9CiAKK3ZhciBpID0gMDsKK3doaWxlIChzdHJpY3RUZXN0 cy5sZW5ndGgpIHsKKyAgICB2YXIgaXRlcmF0aW9uID0gc3RyaWN0VGVzdHMuc2hpZnQoKTsKKyAg ICB2YXIgZGVsYXkgPSBzdHJpY3RUZXN0cy5zaGlmdCgpOworICAgIHRlc3QoaXRlcmF0aW9uLCBk ZWxheSwgY29tcGFyZVN0cmljdCwgInN0cmljdCB0ZXN0ICIgKyArK2kpOworfQorCisvLyBOdW1i ZXIgb2YgY2h1bmtzIHRvIHNlbmQsIGRlbGF5IGJldHdlZW4gY2h1bmtzCiB2YXIgdGhyb3R0bGVk VGVzdHMgPSBbIDUsIDIwLAogICAgICAgICAgICAgICAgICAgICAgICA2LCAzMCBdOwogZnVuY3Rp b24gY29tcGFyZVRocm90dGxlZChjb3VudCwgaXRlcmF0aW9uKQpAQCAtNjEsMTYgKzcwLDExIEBA IGZ1bmN0aW9uIGNvbXBhcmVUaHJvdHRsZWQoY291bnQsIGl0ZXJhdGlvbikKICAgICByZXR1cm4g Y291bnQgPCBpdGVyYXRpb247CiB9CiAKLXRyeSB7Ci13aGlsZSAoc3RyaWN0VGVzdHMubGVuZ3Ro KSB7Ci0gICAgdmFyIGl0ZXJhdGlvbiA9IHN0cmljdFRlc3RzLnNoaWZ0KCk7Ci0gICAgdmFyIGRl bGF5ID0gc3RyaWN0VGVzdHMuc2hpZnQoKTsKLSAgICB0ZXN0KGl0ZXJhdGlvbiwgZGVsYXksIGNv bXBhcmVTdHJpY3QpOwotfQoraSA9IDA7CiB3aGlsZSh0aHJvdHRsZWRUZXN0cy5sZW5ndGgpIHsK ICAgICB2YXIgaXRlcmF0aW9uID0gdGhyb3R0bGVkVGVzdHMuc2hpZnQoKTsKICAgICB2YXIgZGVs YXkgPSB0aHJvdHRsZWRUZXN0cy5zaGlmdCgpOwotICAgIHRlc3QoaXRlcmF0aW9uLCBkZWxheSwg Y29tcGFyZVRocm90dGxlZCk7CisgICAgdGVzdChpdGVyYXRpb24sIGRlbGF5LCBjb21wYXJlVGhy b3R0bGVkLCAidGhyb3R0bGVkIHRlc3QgIiArICsraSk7CiB9CiB9IGNhdGNoKGUpIHsKICAgICBs b2coIkZBSUxFRDogZXhjZXB0aW9uIHJhaXNlZDogIiArIGUubWVzc2FnZSk7Cg== 52483 2010-04-02 23:03:50 -0700 2010-04-02 23:15:12 -0700 Patch bug-36934-20100402230349.patch text/plain 1347 abarth ZGlmZiAtLWdpdCBhL0xheW91dFRlc3RzL0NoYW5nZUxvZyBiL0xheW91dFRlc3RzL0NoYW5nZUxv ZwppbmRleCA5N2U4ZWQ0Li4zMGFhMzI2IDEwMDY0NAotLS0gYS9MYXlvdXRUZXN0cy9DaGFuZ2VM b2cKKysrIGIvTGF5b3V0VGVzdHMvQ2hhbmdlTG9nCkBAIC0xLDUgKzEsMTYgQEAKIDIwMTAtMDQt MDIgIEFkYW0gQmFydGggIDxhYmFydGhAd2Via2l0Lm9yZz4KIAorICAgICAgICBVbnJldmlld2Vk LgorCisgICAgICAgIFJFR1JFU1NJT04oNTYzOTQpOiBodHRwL3Rlc3RzL3htbGh0dHByZXF1ZXN0 L3htbGh0dHByZXF1ZXN0LTUwbXMtZG93bmxvYWQtZGlzcGF0Y2guaHRtbCBmYWlscyBvbiBUaWdl cgorICAgICAgICBodHRwczovL2J1Z3Mud2Via2l0Lm9yZy9zaG93X2J1Zy5jZ2k/aWQ9MzY5MzQK KworICAgICAgICBBdHRlbXB0IEFsZXhleSdzIHNvbHV0aW9uIG9mIHVzaW5nIGEgbm9uLXNuaWZm YWJsZSBNSU1FIHR5cGUuCisKKyAgICAgICAgKiBodHRwL3Rlc3RzL3htbGh0dHByZXF1ZXN0L3Jl c291cmNlcy9kb3dubG9hZC13aXRoLWRlbGF5LnBocDoKKworMjAxMC0wNC0wMiAgQWRhbSBCYXJ0 aCAgPGFiYXJ0aEB3ZWJraXQub3JnPgorCiAgICAgICAgIFVucmV2aWV3ZWQuICBEaXNhYmxlZCBm YXN0L3dvcmtlcnMvd29ya2VyLXRlcm1pbmF0ZS5odG1sIGJlY2F1c2UgaXQKICAgICAgICAgcmFu ZG9tbHkgZmFpbHMgYWJvdXQgMjUlIG9mIHRoZSB0aW1lLgogCmRpZmYgLS1naXQgYS9MYXlvdXRU ZXN0cy9odHRwL3Rlc3RzL3htbGh0dHByZXF1ZXN0L3Jlc291cmNlcy9kb3dubG9hZC13aXRoLWRl bGF5LnBocCBiL0xheW91dFRlc3RzL2h0dHAvdGVzdHMveG1saHR0cHJlcXVlc3QvcmVzb3VyY2Vz L2Rvd25sb2FkLXdpdGgtZGVsYXkucGhwCmluZGV4IDI2OGI1OWYuLjMzZTdmYzggMTAwNjQ0Ci0t LSBhL0xheW91dFRlc3RzL2h0dHAvdGVzdHMveG1saHR0cHJlcXVlc3QvcmVzb3VyY2VzL2Rvd25s b2FkLXdpdGgtZGVsYXkucGhwCisrKyBiL0xheW91dFRlc3RzL2h0dHAvdGVzdHMveG1saHR0cHJl cXVlc3QvcmVzb3VyY2VzL2Rvd25sb2FkLXdpdGgtZGVsYXkucGhwCkBAIC0yLDYgKzIsNyBAQAog aGVhZGVyKCJFeHBpcmVzOiBUaHUsIDAxIERlYyAyMDAzIDE2OjAwOjAwIEdNVCIpOwogaGVhZGVy KCJDYWNoZS1Db250cm9sOiBuby1jYWNoZSwgbm8tc3RvcmUsIG11c3QtcmV2YWxpZGF0ZSIpOwog aGVhZGVyKCJQcmFnbWE6IG5vLWNhY2hlIik7CitoZWFkZXIoIkNvbnRlbnQtVHlwZTogYXBwbGlj YXRpb24veC1uby1idWZmZXJpbmctcGxlYXNlIik7CiAKICRpdGVyYXRpb24gPSAkX0dFVFsnaXRl cmF0aW9uJ107CiAkZGVsYXkgPSAkX0dFVFsnZGVsYXknXTsK