Closed Bug 505031 Opened 17 years ago Closed 17 years ago

Plugins blocklist page can not be opened from addons manager in localized Firefox builds

Categories

(www.mozilla.org :: General, defect, P2)

All
Windows XP

Tracking

(Not tracked)

VERIFIED FIXED

People

(Reporter: unghost, Assigned: clouserw)

References

Details

(Keywords: regression, Whiteboard: [1.9.2-fixed])

Attachments

(2 files)

Attached image Screenshot —
STR: 1. Install Firefox 3.5.1 Russian build for Win32 2. Install some plugin that is blacklisted in Firefox (Quicktime 7.1.3 in my case) 3. Open Tools -> Addons -> Plugins 4. Highlight blocked plugin and click link "More information" Expected results: Firefox shows https://www.mozilla.com/en-US/blocklist/ page Actual results: SSL error dialog is shown with text like "ru.www.mozilla.com uses an invalid security certificate. The certificate is only valid for *.mozilla.com" (see screenshot attached) This error doesn't present in Fx 3.5.1 en-US build. This error doesn't present in Fx 3.0.11 Russian build
Summary: Pluguns blocklist page can not be opened from addons manager in localized Firefox builds → Plugins blocklist page can not be opened from addons manager in localized Firefox builds
This is some problem with the redirecting at mozilla.com. https://www.mozilla.com/ru/blocklist/ is the page the add-ons manager loads which redirects to https://ru.www.mozilla.com/ru/blocklist/ which the certificate doesn't match.
Component: Add-ons Manager → www.mozilla.com
Product: Toolkit → Websites
QA Contact: add-ons.manager → www-mozilla-com
Bug 454299 would fix this.
Depends on: 454299
With the recent WPF plugin blocklisting this is going to hit a whole lot more people. The current behaviour is very malware-y: 1. dialog pops up with scary text 2. more info goes to a security warning page (incidentally, once u ignore the warning, the page is completely useless and doesn't even include the WPF plugin)
We should block on this
Flags: blocking-firefox3.6?
This is also tracked in bug 522864.
Blocks: 522777
Severity: normal → major
Reed, the fun is that https://ru.www.mozilla.com/ru/blocklist/ with a cert exception (or a vaild cert) only redirects again to https://www.mozilla.com/en-US/blocklist/ so we redirect the URL without the locale subdomain the the one with the obsolete locale subdomain to redirect to the en-US page without the locale subdomain. The fix could be "as simple as" changing the redirect for https://www.mozilla.com/*/blocklist/ to go to https://www.mozilla.com/en-US/blocklist/ directly, without redirecting to the obsolete locale subdomain in between. Of course, I don't know the mechanisms behind this where the redirects are set, but saving a useless redirect level and at the same time fixing an instance of bug 454299 and not running into the cert problem would be worth any work on this, I guess :)
(In reply to comment #11) Bug 454299 should fix all that. My testing with the patch makes things go like this: 1. https://de.www.mozilla.com/de/blocklist/ --> 2. https://www.mozilla.com/de/blocklist/ --> 3. https://www.mozilla.com/en-US/blocklist/ We _may_ eventually be able to combine #1 and #2 together, but that would require some heavy dark magic.
As the code doesn't link to the sub-domains at all, the first redirect isn't really crucial, at least not for any current 3.5.x or later. The 2.->3. is the redirect that's golden our lives.
Good, people should only come in with e.g. https://www.mozilla.com/de/blocklist/ so if that redirects in one single step and without going to the locale subdomain in between, we're where we want to be. :)
Flags: blocking-firefox3.6? → blocking-firefox3.6+
Priority: -- → P2
Assignee: nobody → clouserw
Okay, it's nice to know that bug 454299 will fix this issue, but that seems like a bigger fix. Why can't we just stop redirecting people to the locale subdomains right now? The in-product patches link to pages like this: http://www.mozilla.com/en-GB/blocklist/ Seems like we could just show that page... Wil, we're planning on adding a new blocklist that at about 7% of Firefox users have. They will all run into this page. It'd be nice if, tomorrow when the blocklist hopefully gets updated, they don't see the cert error.
Attached patch fix? — — Splinter Review
Any change we do is going to need massive QA, whether it's this bug or the other one - they are both changing the same code. Anyway, here is a patch that removes the redirect. I can't reproduce all the redirects and stuff with SSL in my dev environment. Stuff like https://de.www.mozilla.com/de/mobile/ works fine in production but ends up being a 404 for me, with or without this patch, so it's hard to say if this will work right. Seems to work on www-trunk, maybe we should try this there.
Attachment #409220 - Flags: review?(reed)
Attachment #409220 - Flags: review?(reed) → review+
Comment on attachment 409220 [details] [diff] [review] fix? I also don't have an SSL/non-SSL testbed, so I'll still need to check this on stage. My tests: * http://de.www.example.com/de/firefox/central/ -- no redirect; content displayed * http://de.www.example.com/de/blocklist/ -- 302 redirect to http://www.example.com/en-US/blocklist/ * http://de.www.example.com/blocklist/ -- 301 redirect to http://de.www.example.com/de/blocklist/ * http://www.example.com/de/blocklist/ -- 302 redirect to http://www.example.com/en-US/blocklist/ * http://www.example.com/blocklist/ -- 302 redirect to http://www.example.com/en-US/blocklist/ * http://en-US.www.example.com/en-US/blocklist/ -- no redirect; content displayed * http://www.example.com/en-US/blocklist/ -- no redirect; content displayed All PASS. r=reed
(In reply to comment #20) Oops, I forgot one more test: * http://www.example.com/de/firefox/central/ -- no redirect; content displayed Still PASS.
Status: NEW → ASSIGNED
on stage in r54633. QA: Please check www-trunk after it updates.
Status: ASSIGNED → NEW
I don't have a whole lot of free time tonight to test this, but here's a start (all of these appear to be correct thus far): [1] /firefox: http://de.www-trunk.stage.mozilla.com/de/firefox/ redirects to http://www.mozilla-europe.org/de/firefox/ [2] /whatsnew: https://sv-se.www-trunk.stage.mozilla.com/firefox/3.5.3/whatsnew redirects to https://www-trunk.stage.mozilla.com/sv-SE/firefox/3.5.3/whatsnew/ [3] /first run: https://www-trunk.stage.mozilla.com/firefox/3.5.3/firstrun/ redirects to https://www-trunk.stage.mozilla.com/en-US/firefox/3.5.3/firstrun/ [4] /central: https://www-trunk.stage.mozilla.com/de/firefox/central/ and http://de.www-trunk.stage.mozilla.com/de/firefox/central/ redirect to https://www-trunk.stage.mozilla.com/de/firefox/central/ That's just simple regression testing; would love it if someone could pick the top 5 locales, plus en-US, and test /blocklist, both with and without the locale-subdomain prefix and the /locale/ code too. More tomorrow morning.
I did some locale testing: zh-CN: zh-CN.www-trunk.stage.mozilla.com/zh-CN/blocklist - PASS www-trunk.stage.mozilla.com/zh-CN/blocklist - PASS en-GB: https://en-gb.www-trunk.stage.mozilla.com/en-GB/blocklist -> gives Untrusted Connection ->after accepting the certificate redirects to -> https://www-trunk.stage.mozilla.com/en-US/blocklist/ https://www-trunk.stage.mozilla.com/en-GB/blocklist - PASS fr: https://fr.www-trunk.stage.mozilla.com/fr/blocklist -> gives Untrusted Connection ->after accepting the certificate redirects to -> https://www-trunk.stage.mozilla.com/en-US/blocklist/ https://www-trunk.stage.mozilla.com/fr/blocklist - PASS he: https://he.www-trunk.stage.mozilla.com/he/blocklist -> gives Untrusted Connection ->after accepting the certificate redirects to -> https://www-trunk.stage.mozilla.com/en-US/blocklist/ https://www-trunk.stage.mozilla.com/he/blocklist - PASS es-ES https://es-es.www-trunk.stage.mozilla.com/es-ES/blocklist -> gives Untrusted Connection ->after accepting the certificate redirects to -> https://www-trunk.stage.mozilla.com/en-US/blocklist/ https://www-trunk.stage.mozilla.com/es-ES/blocklist - PASS
They all look fine to me. Is this a=qa?
(In reply to comment #25) > They all look fine to me. Is this a=qa? Let's do it; feeling a lot more confident about this since our testing (although it still makes me slightly nervous, given the magnitude).
(In reply to comment #26) > (In reply to comment #25) > > They all look fine to me. Is this a=qa? > > Let's do it; feeling a lot more confident about this since our testing > (although it still makes me slightly nervous, given the magnitude). If QA is on board, then it works for me. Let's do it...
Thanks everyone, this is on stage and production in r54670.
Status: NEW → RESOLVED
Closed: 17 years ago
Resolution: --- → FIXED
I see it's live and working. :) Big thanks to everyone that made this happen so fast. Great work! :)
Verified FIXED; any fallout goes to a new bug :-)
Status: RESOLVED → VERIFIED
Whiteboard: [1.9.2-fixed]
Component: www.mozilla.org/firefox → www.mozilla.org
Component: www.mozilla.org → General
Product: Websites → www.mozilla.org
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: