Bug#733816: khotkeys launches actions with SIGHUP blocked, leading to unclosable xterms

[Stephen Dowdy <sdowdy@ucar.edu>]

Anthony DeRobertis wrote, On 12/31/2013 02:47 PM:
> On Tue, Dec 31, 2013 at 02:10:42PM -0700, Stephen Dowdy wrote:
> 
>> for i in /proc/[0-9]*/status; do
>>   if sb=$(grep SigBlk ${i}|sed -e 's/^SigBlk:[ \t]//' | grep -v -e '^ffff' -e 00000000000); then
>>     echo "${sb}" "$(cat ${i%/*}/cmdline)";
>>   fi
>> done

I suppose it's a lot easier to type:
    ps -eo blocked,pid,cmd | grep -v \^0000000000000000

;)

> Indeed it appears I might be:
...
> I've got nvidia-kernel-dkms 331.20-2 (from experimental), which is one
> of the affected versions, apparently.

I've backed out to 325.15, successfully.
(no more 80%-of-the-time random 10 second blank screen delays after unlocking the screen,
no more random 50% of the time 30 second no cursor/keyboard input after that,
no more akonadiserver crapping out from spawned mysqlcheck zombie, yay)

> anthony@Zia:~$ grep SigBlk /proc/3289/status 
> SigBlk: 0000000000000001
> 
> so that seems to be sane... except for HUP being ignored. Its parent
> kdeinit4: kded4 [kdeinit] has the same.
> 
> So I'm unsure if this is really bug 728743 or not :-(

I'm betting it is -- thanks for the pointer to that bug #, as i
had not run across it yet.  The nefarious bit of this bug is that
it can create bad behavior in so many things that seem totally
unrelated to the X server/driver, and because the inherited mask
seems somewhat random (memory corruption or failure to dereference
pointer?..), it's just crazy enough to keep you off-track.
(as that bug original started in the 'libgmime' package, was
moved to the "Mono" package before ending up in 'nvidia' ;} )

--stephen

-- 
Stephen Dowdy  -  Systems Administrator  -  NCAR/RAL
303.497.2869   -  sdowdy@ucar.edu        -  http://www.ral.ucar.edu/~sdowdy/