Bug 556286 – www-client/chromium-44.0.2403.89 (and all above?): Segfaults in HTMLParserThread / RawChannel connection broken

• Product: Gentoo Linux
• Status: RESOLVED FIXED

[Description Patrice Levesque 2015-07-30 03:36:12 UTC]

Ever since I emerged =www-client/chromium-44.0.2403.89 the browser when started crashes on every tab, and every extension crashes as well.  Many segfaults are happening.  Browser is hence unusable.  I also tried emerging =www-client/chromium-45.0.2454.15 with the same result.

This is already reported upstream; more details at https://code.google.com/p/chromium/issues/detail?id=505203 but alas no solution.  A workaround is to use '-no-sandbox' as a parameter but this is highly non-recommended.

Reproducible: Always

Steps to Reproduce:
1. Emerge =www-client/chromium-44.0.2403.89
2. Start chromium
3. Look at segfaults in syslog, and for crashed tabs and extensions in chromium.
Actual Results:  
All tabs crash; every extension crashes as well.

Expected Results:  
Not crash.

Might it be possible as a workaround to make older chromium ebuilds available? This started to happen with chromium-44; chromium-43 didn't show that behaviour, and now I cannot easily revert.

[Comment 1 Mike Gilbert 2015-08-03 15:01:56 UTC]

(In reply to Patrice Levesque from comment #0)
> Might it be possible as a workaround to make older chromium ebuilds
> available? This started to happen with chromium-44; chromium-43 didn't show
> that behaviour, and now I cannot easily revert.

You may retrieve old ebuilds from the CVS attic. You do so at your own risk; old versions are known to have un-patched security flaws.

https://sources.gentoo.org/cgi-bin/viewvc.cgi/gentoo-x86/www-client/chromium/?hideattic=0

[Comment 2 Paweł Hajdan, Jr. (RETIRED) 2015-08-13 10:41:33 UTC]

Can you obtain a stack trace using instructions from https://code.google.com/p/chromium/wiki/LinuxDebugging ?

[Comment 3 Paweł Hajdan, Jr. (RETIRED) 2015-09-09 07:13:08 UTC]

See https://code.google.com/p/chromium/issues/detail?id=527209#c7

It seems setting CONFIG_COMPAT_VDSO=n in the kernel config should fix it.

We could add that config check to the ebuild.

[Comment 4 Patrice Levesque 2015-09-12 01:09:49 UTC]

Well done - disabling CONFIG_COMPAT_VDSO in my kernel solved the problem.

Marking as "Resolved / Invalid", as this is not a bug in chromium per se.  Feel free to adjust if needed.

Thanks again,

[Comment 5 Mike Gilbert 2015-09-12 12:16:31 UTC]

Reopening so we remember to add the check to the ebuild.

[Comment 6 Paweł Hajdan, Jr. (RETIRED) 2015-09-20 18:43:57 UTC]

I just pushed a change to chromium.eclass to add the check.

[Comment 7 Mike Gilbert 2015-09-26 15:48:00 UTC]

*** Bug 556006 has been marked as a duplicate of this bug. ***