Closed Bug 1416229 Opened 7 years ago Closed 7 years ago

Firefox NIghtly - no fonts displayed, some square placeholders and question marks instead

Categories

(Core :: Security: Process Sandboxing, defect)

Unspecified
Linux
defect
Not set
normal

Tracking

()

RESOLVED FIXED
mozilla58
Tracking Status
firefox-esr52 --- unaffected
firefox56 --- unaffected
firefox57 --- unaffected
firefox58 --- fixed

People

(Reporter: kkarpieszuk, Assigned: jfkthame)

References

Details

(Keywords: nightly-community, regression)

Attachments

(5 files)

Attached image screenshot from gmail
User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:57.0) Gecko/20100101 Firefox/57.0
Build ID: 20171110002743

Steps to reproduce:

I just updated my Firefox Nightly and on every loaded page I don't see any fonts. I see those weird replacments (please check the screenshot)


Actual results:

No readable fonts, I must fill this bug report using stable Firefox, Nightly is unusable at all


Expected results:

Fonts should be displayed as always
Browser restart does not help
Component: Untriaged → General
Product: Firefox → Core
See Also: → 1404919
Now I opened this bug report with Nightly and I see those wierd fonts but also I see some texts are ok.
Attached image Home page screenshot
I've hit the same issue where no fonts are being displayed, only blocks.

Worked fine until I updated to Nightly.
Looks like this affects Linux users only.
OS: Unspecified → Linux
I have this here also on RHEL-7/x86_64 with "Mozilla/5.0 (X11; Linux x86_64; rv:58.0) Gecko/20100101 Firefox/58.0" (just updated Nightly). 

My screenshot is https://mcepl.fedorapeople.org/tmp/Screenshot%20from%202017-11-10%2014-18-41.png and I have exactly the same on most websites I tried (couple of GitHub and GitLab pages, BBC News, Gnome developer wiki, Red Hat Bugzilla). Surprisingly, https://www.firstthings.com/web-exclusives/2016/01/free-university-orthodoxy and https://duckduckgo.com/ seem to hold together.

I noticed stderr of my Nightly generates a lot of errors:

matej@mitmanek: ~$ firefox

(firefox:28149): GLib-GObject-WARNING **: gsignal.c:3492: signal name 'selection_changed' is invalid for instance '0x7fb698b7de20' of type 'MaiAtkType139'

###!!! [Parent][MessageChannel] Error: (msgtype=0x150084,name=PBrowser::Msg_Destroy) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x2C0091,name=PContent::Msg_StartVisitedQuery) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x150001,name=PBrowser::Msg_AsyncMessage) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x150001,name=PBrowser::Msg_AsyncMessage) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x150001,name=PBrowser::Msg_AsyncMessage) Closed channel: cannot send/recv

--------------------------------------------------------------------

Following advice from bug 1414964 comment 4 I decreased security.sandbox.content.level and although it absolutely didn’t help with fonts but error messages got different.

matej@mitmanek: ~$ firefox

(firefox:28778): GLib-GObject-WARNING **: gsignal.c:3492: signal name 'selection_changed' is invalid for instance '0x7f02031ba7e0' of type 'MaiAtkType139'

(firefox:28778): GLib-GObject-WARNING **: gsignal.c:3492: signal name 'selection_changed' is invalid for instance '0x7f02057f1ab0' of type 'MaiAtkType139'
[Parent 28778, Gecko_IOThread] WARNING: pipe error: Broken pipe: file /builds/worker/workspace/build/src/ipc/chromium/src/chrome/common/ipc_channel_posix.cc, line 709
[Parent 28778, Gecko_IOThread] WARNING: pipe error: Broken pipe: file /builds/worker/workspace/build/src/ipc/chromium/src/chrome/common/ipc_channel_posix.cc, line 709
[Parent 28778, Gecko_IOThread] WARNING: pipe error: Broken pipe: file /builds/worker/workspace/build/src/ipc/chromium/src/chrome/common/ipc_channel_posix.cc, line 709

###!!! [Parent][MessageChannel] Error: (msgtype=0x150084,name=PBrowser::Msg_Destroy) Closed channel: cannot send/recv


###!!! [Child][MessageChannel] Error: (msgtype=0x150001,name=PBrowser::Msg_AsyncMessage) Closed channel: cannot send/recv

matej@mitmanek: ~$
¡Hola KonradK, keslerm!

Could you please try what's suggested on https://bugzilla.mozilla.org/show_bug.cgi?id=1414964#c4 and maybe dupe to that bug?

¡Gracias!
Alex
Flags: needinfo?(kkarpieszuk)
I've hit the same thing on Debian Stretch. I've tried what's suggested in the other bug but it's not making any difference.

I've run through it with mozregression and it's found the following to be at fault:

https://hg.mozilla.org/integration/mozilla-inbound/pushloghtml?fromchange=4ef361678c07c388cdae9c807a7d6f3796ab20fa&tochange=75e7f32c336501a698e618667ab180abc9ff6e84
If this is new in today's Nightly, it is presumably related to bug 1412090. :(  Can someone with an affected system confirm what version of the fontconfig library is present?
I've tried setting the security.sandbox.content.level to 1 and restarting and it had no effect.

I'm seeing the same thing Matěj Cepl was seeing where https://duckduckgo.com/ rendered fine but no other pages worked.

I also was unable to view the source of a website, as it resulted in only boxes as well.

I'm running on a fresh install of Debian 9.
Here is the results from dpkg:


Package: fontconfig
Status: install ok installed
Priority: optional
Section: fonts
Installed-Size: 530
Maintainer: Keith Packard <keithp@debian.org>
Architecture: amd64
Multi-Arch: foreign
Source: fontconfig (2.11.0-6.7)
Version: 2.11.0-6.7+b1
Replaces: fontconfig-config (<< 2.5.93-1)
Depends: libc6 (>= 2.4), libfontconfig1 (>= 2.11), libfreetype6 (>= 2.2.1), fontconfig-config
Pre-Depends: dpkg (>= 1.16.1)
(In reply to Jonathan Kew (:jfkthame) from comment #8)
> If this is new in today's Nightly, it is presumably related to bug 1412090.
> :(  Can someone with an affected system confirm what version of the
> fontconfig library is present?

It's 2.11.0 - or more specifically the debian package version is 2.11.0-6.7+b1.

(In reply to keslerm from comment #9)

> I also was unable to view the source of a website, as it resulted in only
> boxes as well.


I didn't try that, but inspect element showed the right information.
Thanks, folks; that's disturbing, as I thought 2.11.x was plenty new enough. (FWIW, my Ubuntu system has 2.11.94, and fonts are working fine there.

(In reply to keslerm from comment #9)
> I'm seeing the same thing Matěj Cepl was seeing where
> https://duckduckgo.com/ rendered fine but no other pages worked.

Basically, any content that uses webfonts is working for you, but locally-installed fonts are failing.

I'm going to back out the patches from bug 1412090, so tomorrow's Nightly should revert to working again, but I also want to understand what's actually going wrong, given that it works fine on some systems but clearly breaks on others.
Blocks: 1412090
Component: General → Security: Process Sandboxing
Status: UNCONFIRMED → NEW
QA Whiteboard: sb?
Ever confirmed: true
sorry for being late to the party, but as I have been asked:

- sandbox level changing does not fix this for me either

- my fontconfig version is fontconfig 2.11.0-0ubuntu4.2 (I am on Ubuntu 14.04 LTS)
Flags: needinfo?(kkarpieszuk)
Just to wrap up here: this should be fixed in the upcoming Nightly (2017-11-11), as bug 1412090 was backed out yesterday.

The problem turns out (huge thanks to lsalzman for debugging efforts!) to be caused by a bug in certain fontconfig versions (if I'm reading the history correctly, it is present in versions from 2.10.94 to 2.11.1 inclusive): specifically, the serialization/de-serialization of font patterns using FcNameUnparse/FcNameParse is broken in these versions.

We have a workaround that will be included when bug 1412090 is re-landed.

Resolving this as fixed (by backout of the regressing bug).
Status: NEW → RESOLVED
Closed: 7 years ago
Resolution: --- → FIXED
I have a question in general but related: when exactly, at what time Nightlies are pushed to the server? Is it fixed time? 

And thank you for fixing this (I don't have new version yet, but I trust you ;) )
I don't believe there's a fixed time, although I'd expect it to be "pretty soon now". I think it's usually early-ish in the morning (before the start of the working day) according to US Pacific time. Sometimes the builds may take longer than usual, though, or be delayed due to any number of kinds of problems with the infrastructure, etc.; they'll appear whenever they're ready.

(Checking for updates in my Nightly on macOS, I got the 2017-11-11 build already; but I also looked at the linux64 downloads at https://www.mozilla.org/en-US/firefox/nightly/all/ a few minutes ago, and they were still 2017-11-10. So, not quite there yet, but on its way....)
Yes, fontconfig-2.10.95-11.el7.x86_64 and it seems to be fixed with the today's Nightly. Thank you very much.
I confirm it is fixed. Thank you very much!
Assignee: nobody → jfkthame
Target Milestone: --- → mozilla58
> status-firefox57: --- → unaffected

Please, is that certain? 

I ask because in the opening post here <https://bugzilla.mozilla.org/show_bug.cgi?id=1416229#c0> there's: 

>> User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:57.0) Gecko/20100101 Firefox/57.0
>> Build ID: 20171110002743

– and today at <https://mozilla.logbot.info/firefox/20171119#c13878981> we have a comparable screenshot <https://sanqui.keybase.pub/screenshots/2017-11-18T10:37:13-Jul-MozillaFirefox.png> from a user of 57.0.
That looks like bug 1412090.

This specific bug was opened for a temporary issue that occurred with the first attempt to fix bug 1412090, which caused additional breakage for some users (depending on fontconfig version), but was quickly backed out.

Bug 1412090, the original font issue affecting some Linux users in Firefox 57, is on track to be fixed for 59 and currently requesting uplift to 58.

To tell the difference: bug 1412090 affects specific fonts where the user has locally-installed files in "non-standard" locations (e.g. fontconfig has been configured to recognize fonts from a ~/Downloads directory), whereas this issue affected *all* system fonts, the only fonts that worked were those loaded from websites using @font-face.
Thanks for clarification, apologies for the noise.
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: