Re: iceweasel issues (update)

[Michael Cree <mcree@orcon.net.nz>]

On 28/04/12 09:32, Bob Tracy wrote:
> On Fri, Apr 27, 2012 at 12:42:10PM +1200, Michael Cree wrote:
>> I am fairly certain that the crash is this:
>>
>> http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=660960
>>
>> The backtraces I got all ended in madvise() --- which is what the bug  
>> report is about.
> 
> The available facts fit the theory, so I added a mention of what we've
> been seeing on Alpha to the bug report.  Mike's immediate reply
> indicated two things: (1) he's actively working the issue; and (2) the
> underlying cause seems to be someone overlooking the possibility of an
> other-than-4k page size (Alpha has an 8k page size).

Yes, it's a dual issue, namely that the code first assumes an incorrect
4K page size then it does not properly check the return value from
madvise and recalls it again if it did not appear to work.

I have finally got iceweasel version 10.0.6esr-2 to build and have
uploaded to debian-ports unreleased.  I then discovered it still has the
madvise infinite looping problem, and have only just now found out about
a workaround, so feel free to get down the source to iceweasel (make
sure you get version 10.0.6esr-2+alpha) and get the patch from:

https://bug746112.bugzilla.mozilla.org/attachment.cgi?id=616080

modify the test for __powerpc__ to __alpha__, apply it, and build and
try that.

Cheers
Michael.