[gentoo-user] qtwebkit problem (solved)

[gentoo-user] qtwebkit problem (solved)

[Philip Webb]

In order to update to the latest stable PyQt5-5.13.2 ,
I was required to remerge qtwebkit-5.212.0_pre20190629 ;
this normally takes  30 min  & can make its cache in memory.

Towards the end of the latter compile, I suddenly found myself
looking at a completely dark screen : the whole system had crashed.
I pressed the Start button on the box & tried the emerge again,
watching what was going on during the compile,
while monitoring 'free' on another tab & activity on Gkrellm.
Early in the compile, it briefly used most of swap :
I have  8 GB  memory +  8 GB  swap partition
& it was using  c 4,5 GB  swap ; this dropped again fairly quickly.
All  8  processors were running at  100 %  or nearly
& the CPU temperature was  70 C  with fans whining a bit.
Then at the same point in the compile, poof ! -- another hard crash.

Nothing at all in 'syslog' ; 'kern.log' shows :

  Jan 4 13:23:56 localhost kernel: IPv6: ADDRCONF(NETDEV_CHANGE): enp2s0: link becomes ready
  Jan 4 14:09:33 localhost kernel: Kernel logging (proc) stopped.
  Jan 4 14:09:33 localhost kernel: Kernel log daemon terminating.
< this wb the crash, then restart >
  Jan 4 14:09:55 localhost kernel: klogd 1.5.1, log source = /proc/kmsg started.
  Jan 4 14:09:55 localhost kernel: Inspecting /usr/src/linux/System.map
  Jan 4 14:09:55 localhost kernel: Cannot find map file.
  Jan 4 14:09:55 localhost kernel: Loaded 30475 symbols from 1 module.
  Jan 4 14:09:55 localhost kernel: Linux version 4.14.52-gentoo (root@localhost) (gcc version 7.3.0 (Gentoo 7.3.0-r3 p1.4)) #5 SMP Sun Aug 19 23:59:47 EDT 201

which isn't very helpful.  What evidence there was suggested
that it had run out of swap, causing the kernel to panic,
so I added 'dev-qt/qtwebkit notmpfs.conf' to  /etc/portage/package.env ,
tried the emerge again & it went thro' successfully in  29 min .

If you try to emerge LO or FF using a tmpfs , Portage tells you not to.
Rust is the 4th entry in  package.env :
I forget whether Portage/other warned me or I found out the hard way.

It's also a puzzle why qtwebkit failed on this re-merge
after compiling successfully when previously updated a few months ago.

It's a very long time since my system (on any box) crashed in this manner.

This is FYI, but others might have a few comments.

-- 
========================,,============================================
SUPPORT     ___________//___,   Philip Webb
ELECTRIC   /] [] [] [] [] []|   Cities Centre, University of Toronto
TRANSIT    `-O----------O---'   purslowatchassdotutorontodotca

Re: [gentoo-user] qtwebkit problem (solved)

[Peter Humphrey]

On Sunday, 5 January 2020 10:01:23 GMT Philip Webb wrote:

> If you try to emerge LO or FF using a tmpfs , Portage tells you not to.

I just thought I'd say that I haven't had any warnings not to use tmpfs when 
emerging LO or FF. I do have 32GB RAM though, which I dare say makes a 
difference. Maybe portage is even cleverer than I thought!

# grep tmpfs /etc/fstab
tmpfs /var/tmp/portage tmpfs noatime,uid=portage,gid=portage,mode=0775 0 0
tmpfs /tmp tmpfs noatime,nosuid,nodev,noexec,mode=1777 0 0
# ls /etc/portage/*env
/etc/portage/env:
/etc/portage/package.env:

-- 
Regards,
Peter.

Re: [gentoo-user] qtwebkit problem (solved)

[Dale]

Philip Webb wrote:
> In order to update to the latest stable PyQt5-5.13.2 ,
> I was required to remerge qtwebkit-5.212.0_pre20190629 ;
> this normally takes  30 min  & can make its cache in memory.
>
> Towards the end of the latter compile, I suddenly found myself
> looking at a completely dark screen : the whole system had crashed.
> I pressed the Start button on the box & tried the emerge again,
> watching what was going on during the compile,
> while monitoring 'free' on another tab & activity on Gkrellm.
> Early in the compile, it briefly used most of swap :
> I have  8 GB  memory +  8 GB  swap partition
> & it was using  c 4,5 GB  swap ; this dropped again fairly quickly.
> All  8  processors were running at  100 %  or nearly
> & the CPU temperature was  70 C  with fans whining a bit.
> Then at the same point in the compile, poof ! -- another hard crash.
>
> Nothing at all in 'syslog' ; 'kern.log' shows :
>
>   Jan 4 13:23:56 localhost kernel: IPv6: ADDRCONF(NETDEV_CHANGE): enp2s0: link becomes ready
>   Jan 4 14:09:33 localhost kernel: Kernel logging (proc) stopped.
>   Jan 4 14:09:33 localhost kernel: Kernel log daemon terminating.
> < this wb the crash, then restart >
>   Jan 4 14:09:55 localhost kernel: klogd 1.5.1, log source = /proc/kmsg started.
>   Jan 4 14:09:55 localhost kernel: Inspecting /usr/src/linux/System.map
>   Jan 4 14:09:55 localhost kernel: Cannot find map file.
>   Jan 4 14:09:55 localhost kernel: Loaded 30475 symbols from 1 module.
>   Jan 4 14:09:55 localhost kernel: Linux version 4.14.52-gentoo (root@localhost) (gcc version 7.3.0 (Gentoo 7.3.0-r3 p1.4)) #5 SMP Sun Aug 19 23:59:47 EDT 201
>
> which isn't very helpful.  What evidence there was suggested
> that it had run out of swap, causing the kernel to panic,
> so I added 'dev-qt/qtwebkit notmpfs.conf' to  /etc/portage/package.env ,
> tried the emerge again & it went thro' successfully in  29 min .
>
> If you try to emerge LO or FF using a tmpfs , Portage tells you not to.
> Rust is the 4th entry in  package.env :
> I forget whether Portage/other warned me or I found out the hard way.
>
> It's also a puzzle why qtwebkit failed on this re-merge
> after compiling successfully when previously updated a few months ago.
>
> It's a very long time since my system (on any box) crashed in this manner.
>
> This is FYI, but others might have a few comments.
>


Just a thought.  Have you tested the memory with memtest or some other
test tool?  Another thing, maybe rare, power supply can't hold up to the
load?  It's odd that it seems to crash at the same spot as you say,
makes me suspect a bad spot in memory, but it sort of sounds more like a
hardware problem.  Also, is the CPU able to handle those temps?  Mine
never gets that hot.  Of course, I do have a large case which allows for
large coolers.  Mine might hit 50C or so but only when this room is also
pretty warm. 

I'd be tempted to compile another package that will load it down the
same way and see if it crashes then.  If it does, hardware.  If not,
strange. 

Dale

:-)  :-) 

Re: [gentoo-user] qtwebkit problem (solved)

[Dale]

Peter Humphrey wrote:
> On Sunday, 5 January 2020 10:01:23 GMT Philip Webb wrote:
>
>> If you try to emerge LO or FF using a tmpfs , Portage tells you not to.
> I just thought I'd say that I haven't had any warnings not to use tmpfs when 
> emerging LO or FF. I do have 32GB RAM though, which I dare say makes a 
> difference. Maybe portage is even cleverer than I thought!
>
> # grep tmpfs /etc/fstab
> tmpfs /var/tmp/portage tmpfs noatime,uid=portage,gid=portage,mode=0775 0 0
> tmpfs /tmp tmpfs noatime,nosuid,nodev,noexec,mode=1777 0 0
> # ls /etc/portage/*env
> /etc/portage/env:
> /etc/portage/package.env:
>


When emerging some of the packages, FF, LOo and a few others, emerge
runs a precheck for space.  If it doesn't see enough space, it spits out
a warning.  At one point I think it would stop but not sure what it does
now.  Like you, I have 32GBs of ram.  Space for tmpfs isn't much of a
problem.  I can't recall the last time I saw that error message.  Even
when I had 16GBs of memory, it would pass the test.  8GBs may trigger
issues tho. 

The bad thing is, emerge doesn't seem to take into account when emerging
more than one at a time.  If you for example need to upgrade FF and LOo
at the same time, it may have enough for one to compile but not both at
the same time.  With 16GBs in the past, I've had issues with that. 
Since the default is half of memory, it may pass the precheck but run
out of tmpfs when doing both.  For some reason, that happens often.  :/

Dale

:-)  :-) 

Re: [gentoo-user] qtwebkit problem (solved)

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 600 bytes --]

On Sun, 05 Jan 2020 10:43:03 +0000, Peter Humphrey wrote:

> > If you try to emerge LO or FF using a tmpfs , Portage tells you not
> > to.  
> 
> I just thought I'd say that I haven't had any warnings not to use tmpfs
> when emerging LO or FF. I do have 32GB RAM though, which I dare say
> makes a difference. Maybe portage is even cleverer than I thought!

I get warnings for a couple of packages that there is not sufficient
space in $PARTAGE_TMPDIR, which is usually on tmpfs, but they don't
mention tmpfs directly,


-- 
Neil Bothwick

To boldly go where I surely don't belong.

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 833 bytes --]

Re: [gentoo-user] qtwebkit problem (solved)

[Jack]

On 1/5/20 5:01 AM, Philip Webb wrote:
> In order to update to the latest stable PyQt5-5.13.2 ,
> I was required to remerge qtwebkit-5.212.0_pre20190629 ;
> this normally takes  30 min  & can make its cache in memory.
>
> Towards the end of the latter compile, I suddenly found myself
> looking at a completely dark screen : the whole system had crashed.
> I pressed the Start button on the box & tried the emerge again,
> watching what was going on during the compile,
> while monitoring 'free' on another tab & activity on Gkrellm.
> Early in the compile, it briefly used most of swap :
> I have  8 GB  memory +  8 GB  swap partition
> & it was using  c 4,5 GB  swap ; this dropped again fairly quickly.
> All  8  processors were running at  100 %  or nearly
> & the CPU temperature was  70 C  with fans whining a bit.
> Then at the same point in the compile, poof ! -- another hard crash.
>
> Nothing at all in 'syslog' ; 'kern.log' shows :
>
>    Jan 4 13:23:56 localhost kernel: IPv6: ADDRCONF(NETDEV_CHANGE): enp2s0: link becomes ready
>    Jan 4 14:09:33 localhost kernel: Kernel logging (proc) stopped.
>    Jan 4 14:09:33 localhost kernel: Kernel log daemon terminating.
> < this wb the crash, then restart >
>    Jan 4 14:09:55 localhost kernel: klogd 1.5.1, log source = /proc/kmsg started.
>    Jan 4 14:09:55 localhost kernel: Inspecting /usr/src/linux/System.map
>    Jan 4 14:09:55 localhost kernel: Cannot find map file.
>    Jan 4 14:09:55 localhost kernel: Loaded 30475 symbols from 1 module.
>    Jan 4 14:09:55 localhost kernel: Linux version 4.14.52-gentoo (root@localhost) (gcc version 7.3.0 (Gentoo 7.3.0-r3 p1.4)) #5 SMP Sun Aug 19 23:59:47 EDT 201
>
> which isn't very helpful.  What evidence there was suggested
> that it had run out of swap, causing the kernel to panic,
> so I added 'dev-qt/qtwebkit notmpfs.conf' to  /etc/portage/package.env ,
> tried the emerge again & it went thro' successfully in  29 min .
>
> If you try to emerge LO or FF using a tmpfs , Portage tells you not to.
> Rust is the 4th entry in  package.env :
> I forget whether Portage/other warned me or I found out the hard way.
>
> It's also a puzzle why qtwebkit failed on this re-merge
> after compiling successfully when previously updated a few months ago.
>
> It's a very long time since my system (on any box) crashed in this manner.
>
> This is FYI, but others might have a few comments.

There have been other good comments, but I'd ask what are your --jobs 
and --load-average settings?  Does it crash using -j1?  I realize this 
may be moot, since you have already completed the emerge and may not 
want to risk another crash, but it's a thought.

Jack

Re: [gentoo-user] qtwebkit problem (solved)

[Philip Webb]

200105 Jack wrote:
> On 1/5/20 5:01 AM, Philip Webb wrote:
>> In order to update to the latest stable PyQt5-5.13.2 ,
>> I was required to remerge qtwebkit-5.212.0_pre20190629 ;
>> Towards the end of the latter compile, I suddenly found myself
>> looking at a completely dark screen : the whole system had crashed.
> There have been other good comments, but I'd ask what are your --jobs 
> and --load-average settings?  Does it crash using -j1 ?
> I realize this may be moot, since you have already completed the emerge
> and may not want to risk another crash, but it's a thought.

Thanks for this + the other comments.

I'm not interested enough to make any further tests.
I doubt if the problem is hardware, which is  4 yr old  & very reliable ;
the usual rule-of-thumb is 'random = hardware, predictable = software'
& this phenomenon was predictable after the 1st occurence.

Since the previous emerge(s) of this version didn't fail,
the cause has to be some extra demand imposed by the new requirements
eg of the latest PyQt5.  This seems quite likely :
Python + Qt both seem to be projects with little discipline or care,
which keep on bloating & demanding more-more from users.
The Python 2/3 fiasco is a big pain in the neck
& whenever Qt needs updating, you have to delete everything first
before re-installing it, otherwise Portage refuses to handle it.

Back in the days when software had to work on an XT,
programmers had to exercise rigorous self-discipline ;
these days of unlimited memory/storage/processing have made them lazy.
On Gentoo, we get all the advantages of controlling our own systems,
but the downside is that we have to attend to some problems
which are handled by the devs on binary distros.

so back to real life ... (grin)

-- 
========================,,============================================
SUPPORT     ___________//___,   Philip Webb
ELECTRIC   /] [] [] [] [] []|   Cities Centre, University of Toronto
TRANSIT    `-O----------O---'   purslowatchassdotutorontodotca