988490 - Extension block: Blocklist [e-mail] < 2.0.7

Status: RESOLVED FIXED

(Toolkit :: Blocklist Policy Requests, defect)

Description

[Steven Michaud [:smichaud] (Retired)]

This bug is spun off from bug 951906.  See it for more information.

I've been asked to open this bug, so I've done so.  I don't know the details of how blocklisting works.  But I do know that it will be more complicated in this case, because most of the crashes at bug 951906 happen on startup.

Comment 1

[Benjamin Smedberg]

OS: mac only
Addon ID: [e-mail]
Addon version: 2.0.6 and earlier
Firefox version: all versions.

Comment 2

[Jorge Villalobos [:jorgev] (he/him)]

The block is staged now: https://addons-dev.allizom.org/en-US/firefox/blocked/i547

Comment 3

[juan becerra [:juanb]]

I've tested this in staging and the blocklist gets updated with an entry as in comment #2 and the add-on appears disabled.

Comment 4

[Lukas Blakk [:lsblakk] use ?needinfo]

Can we get this up in production in that case?  Will wait for this before sending the 'go' to unthrottle FF28 completely.

Comment 5

[Jorge Villalobos [:jorgev] (he/him)]

It's been pushed to prod now: https://addons.mozilla.org/en-US/firefox/blocked/i568

Comment 6

[kamelboudries5]

pourquoi avoir bloquer 1 pluggins qui me permettai de jouer a un 1 jeu"" happy black jack"" et qui m'est est impossible e reinstaller!! verifier s'il vous plais et debloquer le merci!!

Comment 7

[Steven Michaud [:smichaud] (Retired)]

(In reply to comment #6)

This block should only effect the [e-mail] extension with versions less than 2.0.7.

I sincerely doubt you're using it.  But to be sure check in Tools : Add-ons : Extensions.

Comment 8

[Steven Michaud [:smichaud] (Retired)]

I installed Xunlei for Mac 2.4.0.1356_Beta in /Applications, then installed [e-mail] 2.0.6 (which comes with it) in Firefox 27.  Nothing happened, even after several restarts.

This block may not be working properly.

Next I'll try with FF 28.

Comment 9

[Steven Michaud [:smichaud] (Retired)]

I updated FF 27 to FF 28 in About Firefox, and it made no difference.  This block still has no effect.

Comment 10

[Jorge Villalobos [:jorgev] (he/him)]

Juan, can you check if you see different behavior on the test block and the production block?

Comment 11

[Steven Michaud [:smichaud] (Retired)]

Ping.

This really does need to be addressed.  The crashes at bug 951906 are still (currently) the #3 Mac topcrasher on the 28 branch and the #4 Mac topcrasher on the 27 branch.

Comment 12

[Jorge Villalobos [:jorgev] (he/him)]

(In reply to Steven Michaud from comment #11)
> Ping.
> 
> This really does need to be addressed.  The crashes at bug 951906 are still
> (currently) the #3 Mac topcrasher on the 28 branch and the #4 Mac topcrasher
> on the 27 branch.

I'm trying to test this myself but I haven't been able to download the DMG that causes the crash. Can you share it somehow?

The only thing I can think might be breaking this block is that it is restricted for OS X (os="Darwin"). That means it shouldn't have worked on staging either, though. If this is sufficiently critical, we can remove the restriction and see if that works...

Blair, is the "os" attribute in extension blocks handled differently than for plugins? Is "Darwin" the right value to set if I want an OS X-only block?

Comment 13

[Steven Michaud [:smichaud] (Retired)]

Here's a link, which seems to work, to the last crashing version:

http://down.sandai.net/mac/thunder_dl2.4.0.1356_Beta.dmg

Comment 14

[Steven Michaud [:smichaud] (Retired)]

And here's a link to the first non-crashing version:

http://down.sandai.net/mac/thunder_dl2.5.0.1408_Beta.dmg

There's now also a 2.6.0 beta, which appeared since the last time I looked:

http://down.sandai.net/mac/thunder_dl2.6.0.1440_Beta.dmg

Comment 15

[Blair McBride [:Unfocused] (UNAVAILABLE)]

(In reply to Jorge Villalobos [:jorgev] from comment #12)
> Blair, is the "os" attribute in extension blocks handled differently than
> for plugins? 

They're handled the same.

> Is "Darwin" the right value to set if I want an OS X-only block?

Yes.

Comment 16

[Jorge Villalobos [:jorgev] (he/him)]

(In reply to Steven Michaud from comment #13)
> Here's a link, which seems to work, to the last crashing version:
> 
> http://down.sandai.net/mac/thunder_dl2.4.0.1356_Beta.dmg

I've been trying that URL, but it only downloads a few Kb and then quits on its own. If you can download the DMG, can you share it somewhere?

Comment 17

[Steven Michaud [:smichaud] (Retired)]

Weird.  I suspect something/someone is blocking your access to that site or those files.

I've uploaded all the Thunder DMGs I could find to my people.mozilla.com account:
http://people.mozilla.org/~stmichaud/xunlei/

Let us know your results.

Comment 18

[Steven Michaud [:smichaud] (Retired)]

> all the Thunder DMGs I could find

Just to make it clear:  These all came from the http://down.sandai.net/ site, via links on the following page (in Chinese):  http://mac.xunlei.com/history_version.html.

I didn't just round them up from random locations on the web :-)

Comment 19

[Jorge Villalobos [:jorgev] (he/him)]

I just installed version 2.0.6, and the production block seems to be working correctly. Tested this on a recent Aurora version.

Comment 20

[Steven Michaud [:smichaud] (Retired)]

On Robert Strong's advice, I just tested again with FF 28, and also found it working correctly.

What I did differently this time is let FF 28 sit for half an hour -- the last time I only let it sit for about 5 minutes.  When I looked again at the VM in which it was running, I found an "Add-ons may be causing trouble" dialog had popped up.  It prompted me to disable "Thunder extension 2.0.6" and restart Firefox.  When I restarted the extension was indeed disabled.

I don't know exactly how long it took for the dialog to appear.  I'll test again to try to find out.

But in any case this kind of block won't work for the majority of those seeing bug 951906's crashes, since they mostly happen on startup.  So (as I understand it) we'll also need to block [e-mail] < 2.0.7 in the "built-in block list".  The "built-in blocklist" can get updated by the updater.  But for many the crashes happen too soon for them to have time to look for an update.  So we won't be able to completely fix these crashes until the "built-in blocklist" containing a block for [e-mail] < 2.0.7 gets included in a release (which many will need to download using a different browser).

(*All* the crashes happen after the [e-mail] extension is loaded, but it gets loaded just seconds after FF starts up.)

Comment 21

[Steven Michaud [:smichaud] (Retired)]

I deleted ~/Library/Application Support/Firefox/, ~/Library/Application Support/Mozilla/ and ~/Library/Caches/Firefox/ and re-ran my test.  FF 28 sat for 10 minutes before a dialog appeared offering to block "Thunder extension 2.0.6".

Comment 22

[Steven Michaud [:smichaud] (Retired)]

Shall we just close this bug again, since the (non-built-in) block seems to be working correctly?

Note that bug 951906 will need to remain open for a while ... at least (I'd say) until we have a release whose built-in blocklist contains the block for [e-mail] < 2.0.7.

Comment 23

[Steven Michaud [:smichaud] (Retired)]

(Following up comment #21)

I just ran the same test with FF 28 and thunder_dl2.6.0.1440_Beta.dmg (which contains v 2.0.7 of the [e-mail] extension):  No crashes, and no block (in other words no problems).

Comment 24

[Steven Michaud [:smichaud] (Retired)]

Should I open a bug on getting the [e-mail] < 2.0.7 block into the built-in blocklist?

Comment 25

[Jorge Villalobos [:jorgev] (he/him)]

The blocklist is updated once a day, so it can take a while. If you want to force a blocklist reload, you can follow these instructions: https://wiki.mozilla.org/Blocklisting/Testing.

The blocklist that ships with Firefox should include the latest blocks, though that's something that probably should be consulted with the build folks, since I don't know the exact timing for this.

Comment 26

[Steven Michaud [:smichaud] (Retired)]

> probably should be consulted with the build folks

I want to do this as soon as possible, but I don't know who to talk to.

If you know, please needinfo the appropriate person/people.  Otherwise let me know (here) and I'll start the needinfo hunt on my own.

Comment 27

[Jorge Villalobos [:jorgev] (he/him)]

Lukas, do you know if every build includes the latest version of the blocklist? Or is it cached somehow?

Comment 28

[Lukas Blakk [:lsblakk] use ?needinfo]

(In reply to Jorge Villalobos [:jorgev] from comment #27)
> Lukas, do you know if every build includes the latest version of the
> blocklist? Or is it cached somehow?

Every build does include the latest version (as we have automated updates being pushed to the tree) but if someone's profile already has a blocklist in it, they can only get updated by the downloaded blocklist (daily) until bug 639524 is fixed.

Comment 29

[Steven Michaud [:smichaud] (Retired)]

Thanks, Lukas!

So we don't have to do anything special to get this block into the next FF release's built-in blocklist.  But bug 639524 is going to prevent most of bug 951906's victims from seeing any relief until *it* gets into a release.  So bug 639524 in effect blocks a Mac topcrasher, and becomes seriously more urgent.

I'll add a comment to bug 639524 to indicate this, and request that it be tracked.