1169113 - crash in libType1Scaler.dylib@0x161f67 | CFRunLoopRunSpecific

Status: RESOLVED INVALID

(Core :: Widget: Cocoa, defect)

Description

[Kartikaya Gupta (email:[e-mail])]

This bug was filed from the Socorro interface and is 
report bp-514bba0a-895c-44fd-895a-0d15b2150528.
=============================================================

I've been hitting this crash a lot on OS X 10.6.8 (yes, old OS X). No real STR; mostly seems to happen while Firefox is in the background.

bp-6ffb8b22-b69b-43f8-b463-46f702150520
bp-0d5bc04a-ac9f-459c-8076-404f82150520
bp-da8fcc8b-71c9-408b-a711-182902150502
bp-2dd464d9-15ab-4d67-afcb-6a3212150523
bp-5abc433c-bca0-4e4e-8f11-db71f2150528

Comment 1

[Steven Michaud [:smichaud] (Retired)]

That a crash happens in CFRunLoopSpecific tells us hardly anything at all.  But some of your crashes are in libType1Scaler.dylib, which tells us a bit more -- they have something to do with fonts.

So ...

Have you installed any custom fonts on your 10.6.8 box?  Have you changed any font-specific settings in Firefox?

Jonathan, do you have any ideas what may be going on here?

Comment 2

[Jonathan Kew [:jfkthame]]

libType1Scaler suggests there may be a legacy PostScript font involved, perhaps one of dubious quality... I wouldn't expect to see this used for any of the (TrueType) fonts that Apple ships with the OS.

So checking your Library/Fonts folders (or Font Book.app) for old fonts that you don't really need and removing them, at least temporarily, might help to narrow things down.

Comment 3

[Kartikaya Gupta (email:[e-mail])]

(In reply to Steven Michaud from comment #1)
> Have you installed any custom fonts on your 10.6.8 box?

I don't recall doing so explicitly but it's possible I did so and forgot.

>  Have you changed
> any font-specific settings in Firefox?

Not that I know of. I went to about:config and the only prefs with "font" in the name that show up as changed are:
font.internaluseonly.changed -> set to false
print.print_downloadfonts -> set to false
print_downloadfonts -> set to false

(In reply to Jonathan Kew (:jfkthame) from comment #2)
> libType1Scaler suggests there may be a legacy PostScript font involved,
> perhaps one of dubious quality... I wouldn't expect to see this used for any
> of the (TrueType) fonts that Apple ships with the OS.
> 
> So checking your Library/Fonts folders (or Font Book.app) for old fonts that
> you don't really need and removing them, at least temporarily, might help to
> narrow things down.

Font Book lists 107 fonts on my machine, and the validation tool in Font Book says all of them are good. Most of them are .ttf but I see a few .ttc, .dfont, and .otf files listed amongst them. I can disable the non-English ones to start (which I'm more confident I don't think I need) and see if that helps.

Comment 4

[Jonathan Kew [:jfkthame]]

I'm not very optimistic this'll help, as it doesn't sound like you have a lot of old junk fonts lying around.

It's also possible that the stack frame is spurious, perhaps due to some kind of earlier corruption, and the crash has nothing to do with fonts at all... dunno, really.

Comment 5

[Steven Michaud [:smichaud] (Retired)]

> It's also possible that the stack frame is spurious

So it is.

Offset 0x161f67 in OS X 10.6.8's libType1Scaler.dylib is a meaningless address in its __const section.  It's an impossible value for a crash address.

So until and unless we have STR that are close to 100% effective, there's probably nothing we can do.

Furthermore, there seem to be very few of these crashes in Socorro, at either address.

I'm going to resolve this INCOMPLETE.  Please reopen it if you find good STR.

Comment 6

[Steven Michaud [:smichaud] (Retired)]

Kats, here's one thing you might do short of providing good STR, which would be very helpful:

Run a mozilla-central nightly in gdb until you see one of these crashes, then get an all-thread stack trace ("thread apply all bt"), then attach that trace here.  It needs to be an m-c nightly, because only those don't have their symbols stripped.

Comment 7

[Steven Michaud [:smichaud] (Retired)]

Note that if you're running in e10s mode, you'll need to run your m-c nightly and then attach (with gdb) to both your main process and your content process.

Comment 8

[Steven Michaud [:smichaud] (Retired)]

Yet another note:

OS X (all client versions) can get pretty flaky if you let it run without rebooting for much longer than a day.  If that's you, then try rebooting -- you may find that your crashes vanish into thin air.

Comment 9

[Kartikaya Gupta (email:[e-mail])]

Attachment: Thread backtraces

I ran the latest nightly with gdb attached and eventually hit another crash in the main process. I'm attaching the thread backtraces. Again not much in the way of STR except this time Nightly was in the foreground and I had just clicked on a link. The tab spinner started up and that's when it happened.

And my uptime is ~46 days so I guess it might be worth a reboot.

Comment 10

[Kartikaya Gupta (email:[e-mail])]

Here's the crash report corresponding to comment 9

https://crash-stats.mozilla.com/report/index/92c3b40f-aef7-4253-902b-297a92150529

Comment 11

[Steven Michaud [:smichaud] (Retired)]

Thanks for the gdb stack trace.  Unfortunately it's not any more informative than your Socorro stacks.

I await the results of your rebooting :-)

Comment 12

[Kartikaya Gupta (email:[e-mail])]

Not conclusive yet, but since the reboot I haven't experienced any of these crashes.