Closed
Bug 800708
Opened 12 years ago
Closed 10 years ago
Test failure "All downloads have been finished" in /testPrivateBrowsing/testPrivateDownloadPanel.js and in /testPrivateBrowsing/testDownloadManagerClosed.js
Categories
(Mozilla QA Graveyard :: Mozmill Tests, defect, P2)
Mozilla QA Graveyard
Mozmill Tests
Tracking
(firefox25 wontfix, firefox26 disabled, firefox27 disabled, firefox-esr10 wontfix, firefox-esr17 wontfix)
People
(Reporter: AndreeaMatei, Unassigned)
References
()
Details
(Whiteboard: [mozmill-test-failure])
Attachments
(3 files, 3 obsolete files)
This happened today on Windows Vista x86, with ERS branch 10.0.9. It seems the download took very long so this waitFor has failed: // Wait until all downloads have been finished controller.waitFor(function () { return dm.activeDownloadCount === 0; }, "All downloads have been finished"); I will continue to monitor this over the weekend.
Reporter | ||
Updated•12 years ago
|
Comment 1•12 years ago
|
||
Please add a link to the failure list and not to a single report. Thanks.
Priority: -- → P2
Reporter | ||
Updated•12 years ago
|
Comment 2•12 years ago
|
||
It's happening for releases on esr10. Lets get this figured out and fixed.
Whiteboard: s=121022 u=failure c=private_browsing p=1
Reporter | ||
Updated•12 years ago
|
Assignee: nobody → andreea.matei
Status: NEW → ASSIGNED
Reporter | ||
Updated•12 years ago
|
status-firefox18:
--- → affected
status-firefox19:
--- → affected
Reporter | ||
Comment 3•12 years ago
|
||
This it's not reproducing for me and I also haven't seen it fail since October 13th.
Comment 4•12 years ago
|
||
So it appears that this has occurred only six times since June. Four of these are on Windows NT 6.0.6002 and all of these were on different versions of Firefox. Given how infrequent this occurs I might suggest adding some debug to the test so that when it next occurs we can find out what's still downloading. The files are locally hosted so the downloads should be very quick.
Reporter | ||
Comment 5•12 years ago
|
||
I got it to reproduce, like one time on 300 runs. I found out that in the current waitFor, when we fail, the first 4500ms the activeDownloadCount = 2, then for the next 5000 is 1. Somehow it's not enough time to get to none actively downloading.
Comment 6•12 years ago
|
||
It's great that you're able to reproduce this! The downloads should be quick because they're served locally, but maybe there is some delay in starting to download them. I wonder if we should either increase the timeout, or have additional waits. For example, we could wait the default amount of time for each download rather than for all downloads.
Comment 7•12 years ago
|
||
Andrea, can you please increase the timeout and check how long it takes for the other remaining files to download? I wonder if the initial download is only affected here.
Reporter | ||
Comment 8•12 years ago
|
||
When increased the timeout (even until 15000), that waitFor passes, but fails later at the dm.close(). Need to investigate further to see if that happens in pb mode or right at the end. It's hard to catch it, today I runned tests all day on vista VM. Still, the first download takes the longest time.
Comment 9•12 years ago
|
||
Please use as many as possible dump statements which also print the current time to the console. So we can see what's happening.
Reporter | ||
Comment 10•12 years ago
|
||
Looks like is at the last line on the test. dm.close(); This should have to be the other failure from the reports. I'm checking now what timeout is the most appropriate for the two waitFor() out and in PB mode. But is always the first file that takes a lot longer to download (unknown_type.mtdl)
Comment 11•12 years ago
|
||
I kinda would like to see some console output for this problem. Please attach it to the bug.
Reporter | ||
Comment 12•12 years ago
|
||
This is the output for the "All downloads have been finished" error, outside PB. Dump doesn't work on the windows console, so I send the output to a txt file. There's also another case where in the first waitFor it takes long, but manages to pass and in the second one inside PB fails (got that even with 8000 timeout).
Reporter | ||
Comment 13•12 years ago
|
||
This is the output for "Download Manager has been closed" error. Used dump before and after the last close of download manager and only the one before appears.
Comment 14•12 years ago
|
||
(In reply to Andreea Matei [:AndreeaMatei] from comment #13) > Created attachment 679119 [details] > DownloadManagerClose Error That one should be attached to bug 802126.
Comment 15•12 years ago
|
||
(In reply to Andreea Matei [:AndreeaMatei] from comment #12) > Created attachment 679117 [details] > waitForTimeout Error This does not contain useful information. Dump states with timestamps as I already have mentioned in comment 6 yesterday.
Reporter | ||
Comment 16•12 years ago
|
||
Here are the timestamps. I've used 8000 timeout, for the first one it passed at 5600 and the second in PB mode failed. If that's the case, I feel that separating the downloads the first time, with waitFor() for each of them will not solve the issue in PB mode where is only one file.
Attachment #679117 -
Attachment is obsolete: true
Comment 17•12 years ago
|
||
Again please add more context. Dump which file its starting to download and when it is ready. As best log each step, which would give us really helpful information.
Reporter | ||
Comment 18•12 years ago
|
||
Both files are starting at the beginning in the for, I will add a dump for which one is finished first. In PB mode where also fails is just one file: unknown_type_pb.stbf.
Reporter | ||
Comment 19•12 years ago
|
||
So here are the results, showing each download file, it's state and the timestamp. state 7 means scanning according to downloads.js and state 1 is finished. Still, I have another one where both are still downloading at 8000. Even if we increase timeout even further, it's odd it takes that long. I think we could indeed separate the downloads, start the next one only when the first is done and wait for each of them. Couldn't get it to fail like this.
Attachment #679119 -
Attachment is obsolete: true
Attachment #679658 -
Attachment is obsolete: true
Comment 20•12 years ago
|
||
Do you have an antivirus software installed on your box? That could be the cause. Our CI boxes don't have. If that's the case I wonder if we should try to disable the scan for our test files.
Reporter | ||
Comment 21•12 years ago
|
||
This is still failing sometimes (1 in 20 runs) with just downloading each file separately, with default timeout. If we increase timeout, even double it, it stops failing, but this is not a wanted fix since it shouldn't be taking this long to download a local file, which indicates a possible firefox bug.
Comment 22•12 years ago
|
||
Failing under which conditions? Normal CPU load? The VM shouldn't have a high load usually, so I would say you should test it directly on it which probably gives us better results.
Reporter | ||
Comment 23•12 years ago
|
||
I've tested under high load CPU, but I will check this morning how it behaves with normal load. Thanks! I assume it will be fine by downloading the files synchronously.
Reporter | ||
Comment 24•12 years ago
|
||
Downloading each file and waiting for it to finish.
Attachment #682403 -
Flags: review?(hskupin)
Attachment #682403 -
Flags: review?(dave.hunt)
Comment 25•12 years ago
|
||
Comment on attachment 682403 [details] [diff] [review] patch v1 Review of attachment 682403 [details] [diff] [review]: ----------------------------------------------------------------- Before fully reviewing this patch I would like to know how the downloading times are looking now? You haven't given any numbers for this patch. ::: tests/functional/testPrivateBrowsing/testDownloadManagerClosed.js @@ +67,4 @@ > downloads.downloadFileOfUnknownType(controller, DOWNLOADS[i]); > + > + // Save information of currently downloaded file > + downloadedFiles = dm.getAllDownloads(); I think that this should not be called within the loop. if it also returns already downloaded files we are good, otherwise we will leave files around on the file system.
Attachment #682403 -
Flags: review?(hskupin)
Attachment #682403 -
Flags: review?(dave.hunt)
Attachment #682403 -
Flags: review-
Reporter | ||
Comment 26•12 years ago
|
||
This is the highest report, 600 ms for one file. As you can see in the file, the getAllDownloads() method returns those already downloaded as well (when the first is done, is still kept in downloadedFiles with the currently downloading one).
Comment 27•12 years ago
|
||
Wow, this smells like a real bug in Firefox! Can you please exchange the files we download from local ones to some remote files? If that also takes that long from state 7 to state 1 we should deeper investigate it and file a bug against Firefox.
Reporter | ||
Comment 28•12 years ago
|
||
With remote files (the same, from mozqa.com), in 20 runs, mostly they were also about 600ms, one case with 900ms and 1200ms but also one failed: First download reached the timeout (5000), second download started and was finished at 1500 while the first was still downloading.
Comment 29•12 years ago
|
||
The important thing here is the transition from state 7 to state 1. It's not relevant how long it takes to download the files. So can you give me that information for remote files being downloaded synchronously and asynchronously?
Reporter | ||
Comment 30•12 years ago
|
||
For remote files, downloading separately, the transition in as shown in the file attached, state 7 persists until it's 1. Downlowding both in the for loop, we have the following: * states 7 and 1: http://pastebin.mozilla.org/1960791 - one gets downloaded very quickly and the other stays at 7 for a time, then gets to 1 * states 5 and 1: http://pastebin.mozilla.org/1960789 - 5 stands for "The download is in the queue but is not presently downloading." and is strange that is not changing to another state but suddenly into 1 * states 1 and 0: http://pastebin.mozilla.org/1960840 - 0 stands for "The download is in the process of being downloaded." - one file is downloaded quick, the other is 0 from the beginning
Comment 31•12 years ago
|
||
(In reply to Andreea Matei [:AndreeaMatei] from comment #30) > * states 7 and 1: http://pastebin.mozilla.org/1960791 - one gets downloaded > very quickly and the other stays at 7 for a time, then gets to 1 Given that pastebin links make it hard for us to query for and get obsolete please always paste content in the bug. I would prefer a combined list of events not separated by type. Please attach such a file. Something is wrong here downloaded Files: [object Object],[object Object] active download count: 0 download http://mozqa.com/data/firefox/downloading/unknown_type.fmtd state 1 download http://mozqa.com/data/firefox/downloading/unknown_type.mtdl state 7 Why does it give 0 for active download count while there is still one download in progress? > * states 5 and 1: http://pastebin.mozilla.org/1960789 - 5 stands for "The > download is in the queue but is not presently downloading." and is strange > that is not changing to another state but suddenly into 1 As I have suspected above it probably takes a long time to initialize the scanning module. For follow-up downloads it's way shorter. But why does that only happen for parallel downloads? That's something we have to figure out. > * states 1 and 0: http://pastebin.mozilla.org/1960840 - 0 stands for "The > download is in the process of being downloaded." - one file is downloaded > quick, the other is 0 from the beginning Not sure how I should read it because I miss the overall picture of that test. Please provide the full log as mentioned above.
Updated•12 years ago
|
Assignee: andreea.matei → andrei.eftimie
Comment 32•11 years ago
|
||
Happened on Windows 8 x86 and ESR17 branch: http://mozmill-ci.blargon7.com/#/functional/report/72e15dc943833b8fcba70aeb516fa7dd
status-firefox-esr17:
--- → affected
Updated•11 years ago
|
status-firefox20:
--- → affected
status-firefox21:
--- → affected
Comment 33•11 years ago
|
||
Have not had any more failures in a full month: http://mozmill-ci.blargon7.com/#/functional/top?branch=All&platform=All&from=2013-01-25&to=2013-02-25 This test has been refactored as part of bug 823064 (applied to all branches except esr17 and release). I'd say we should close it for now and reopen if it still occurs.
Comment 34•11 years ago
|
||
Sounds good. Please close it yourself next time.
Status: ASSIGNED → RESOLVED
Closed: 11 years ago
Resolution: --- → WORKSFORME
Reporter | ||
Comment 35•11 years ago
|
||
Happened again on Win 8, with 17.0.4esrpre: http://mozmill-ci.blargon7.com/#/functional/report/2a6536e9db9f5f44ed48c58510585b14
Status: RESOLVED → REOPENED
Resolution: WORKSFORME → ---
Comment 36•11 years ago
|
||
Same issue happening on the rewritten test (actual download code is the same): http://mozmill-ci.blargon7.com/#/functional/report/db924715258381629c26ddbad6e6eb29
Updated•11 years ago
|
Summary: Test failure "All downloads have been finished" in /testPrivateBrowsing/testDownloadManagerClosed.js → Test failure "All downloads have been finished" in /testPrivateBrowsing/testPrivateDownloadPanel.js and in /testPrivateBrowsing/testDownloadManagerClosed.js
Reporter | ||
Comment 37•11 years ago
|
||
Happened again on Win 8, but on Aurora now: http://mozmill-ci.blargon7.com/#/functional/report/c8094a822ef568b588c5eddaca32882f
Updated•11 years ago
|
Status: REOPENED → ASSIGNED
status-firefox18:
affected → ---
status-firefox19:
affected → ---
status-firefox22:
--- → affected
status-firefox23:
--- → affected
Comment 38•11 years ago
|
||
Happened again on Windows Vista with Aurora DE: http://mozmill-ci.blargon7.com/#/functional/report/8aa237803fd4b0e7348adc8bbe54ab1d
Comment 39•11 years ago
|
||
Happened today on Windows 8 x64 with Nightly en-US build: http://mozmill-ci.blargon7.com/#/functional/report/206c36485a655b168e74264237372d2e
status-firefox25:
--- → affected
OS: Windows Vista → All
Reporter | ||
Comment 40•11 years ago
|
||
Has reproduced again on Windows 7, with Aurora: http://mozmill-daily.blargon7.com/#/functional/report/180cf2548ef2865af3ae441d61c52967
Reporter | ||
Updated•11 years ago
|
Whiteboard: s=121022 u=failure c=private_browsing p=1 → [mozmill-test-failure] s=121022 u=failure c=private_browsing p=1
Comment 41•11 years ago
|
||
Happened today on Windows 7 with Firefox 25.0a2 (25.0a2, en-US, 20130807004004) http://mozmill-daily.blargon7.com/#/functional/report/bccc2ea9ac1e478fa074c699d8bb01be
Updated•11 years ago
|
Whiteboard: [mozmill-test-failure] s=121022 u=failure c=private_browsing p=1 → [mozmill-test-failure]
Comment 42•11 years ago
|
||
Happened again on WinXP with Firefox 26.0a1 (26.0a1, de, 20130831030224): http://mozmill-daily.blargon7.com/#/functional/report/3c3cd991de01b704f3f65783a18850b6
status-firefox26:
--- → affected
Comment 43•11 years ago
|
||
That failure is not related since the test is skipped. Might be a machine problem?
Comment 44•11 years ago
|
||
Agreed, I have no idea why this report have been filed as comment here. It's clearly not related also because it's a disconnect and not the failure as covered on this bug.
Reporter | ||
Comment 45•11 years ago
|
||
We no longer have testDownloadManagerClosed.js in other branches besides esr17 and testPrivateDownloadPanel.js is skipped on aurora and default due to the Downloads API changes. Still fails on Beta with Windows 8 intermittently.
status-firefox20:
affected → ---
status-firefox21:
affected → ---
status-firefox22:
affected → ---
status-firefox23:
affected → ---
status-firefox27:
--- → disabled
Updated•10 years ago
|
Assignee: andrei.eftimie → nobody
Status: ASSIGNED → NEW
Comment 46•10 years ago
|
||
This failed again on ESR24: http://mozmill-daily.blargon7.com/#/functional/report/2f982f72826307fed840a3b11cbd121d Due to the fact that on all newer branches we'll have the new Download Library implementation, this will not fail _probably_ the same way. We won't fix this for ESR24, so I'm closing this issue now. If something similarly fails after bug 908649 lands, please file a different bug.
Status: NEW → RESOLVED
Closed: 11 years ago → 10 years ago
Resolution: --- → WONTFIX
Updated•5 years ago
|
Product: Mozilla QA → Mozilla QA Graveyard
You need to log in
before you can comment on or make changes to this bug.
Description
•