/usr/lib/libnss3.so: version `NSS_3.19.1' not found

[YuGiOhJCJ Mailing-List]

On Sun, 29 May 2016 13:17:14 +0100
Wols Lists <antlists at youngman.org.uk> wrote:

> On 29/05/16 09:21, YuGiOhJCJ Mailing-List wrote:
> > On Sun, 29 May 2016 01:35:47 +0100
> > Wols Lists <antlists at youngman.org.uk> wrote:
> > 
> >> On 24/05/16 11:26, Eike Rathke wrote:
> >>> Hi YuGiOhJCJ,
> >>>
> >>> On Thursday, 2016-05-19 17:26:21 +0200, YuGiOhJCJ Mailing-List wrote:
> >>>
> >>>>> Dumb question: how much system memory is available?
> >>>> I have 4GB of memory:
> >>>
> >>> That certainly is not enough and it will either grind your machine to
> >>> heavily swap, or break the build / abort things if no swap is available.
> >>>
> >>>> $ free -m
> >>>>              total       used       free     shared    buffers     cached
> >>>> Mem:          3995        559       3436          0         53        330
> >>>> -/+ buffers/cache:        175       3819
> >>>> Swap:          956          0        956
> >>>
> >>>
> >>>
> >>>>> And why are you building under /tmp/ and how much free disk space is
> >>>>> there?
> >>>>
> >>>> Well, I could do it in /home but as it is a NFS share, it is slower than in /tmp
> >>>
> >>> Ok, but as Linoel already said, using /var/tmp/ might be a better
> >>> choice. Also, if disk space is limited under /tmp/ then building there
> >>> may conflict with temporary files the compiler and linker create, which
> >>> can become quite large.
> >>
> >> Bear in mind, the LFS says that /tmp and /var/tmp behave differently. On
> >> a "correctly" configured system, the contents of /tmp are NOT guaranteed
> >> to survive a system crash. Which is why /tmp is often configured as a
> >> tmpfs. On the other hand, the contents of /var/tmp ARE guaranteed to
> >> survive, which is why vi and emacs and that sort of program all store
> >> their replay logs there ...
> >>
> >> and which is why the OP's choice of /tmp was probably correct :-)
> >> although most distros don't seem to make the /tmp directory overly
> >> large. (They also seem not to allocate much swap space.)
> >>>
> >>>> Do you think I don't have enough memory?
> >>>> Is there a way to require less memory while building libreoffice or should I buy more memory?
> >>>
> >>> Buy memory ;-)  at least 8GB are needed, but when building with debug
> >>> and symbols even that might result in swapping if you forgot to quit
> >>> a previous gdb session before linking Calc for example.. 12GB or having
> >>> a larger swap than just 1GB is recommended.
> >>>
> >> My rule of thumb is simple. Disk space is cheap, I allocate twice
> >> maximum ram per disk. In other words, my desktop is maxed out at 16Gb so
> >> the two disks each have a 32Gb swap partition. My laptop maxes out at
> >> 8Gb so there should be a 16Gb swap partition on the drive (actually it's
> >> 32Gb :-)
> >>
> >> The reason for that is - in the old days everybody said "swap should be
> >> twice memory" which was thought to be an old wives' tale. Then kernel
> >> 2.4 came out, and it turned out (1) that this requirement was actually
> >> part of the swap algorithm, and (2) the optimisations and hacks and
> >> whatever that enabled smaller swaps were a heap of old crufty rubbish.
> >> Linus ripped out all the hacks and vanilla 2.4 kernels started crashing
> >> everywhere they had a swapspace of less than twice ram.
> >>
> >> Obviously, new optimisations have gone in, presumably much better than
> >> before, but nowhere have I found any reference to whether the
> >> fundamental algorithm has been replaced. So I'm assuming it hasn't, and
> >> allocate at least twice ram to ensure I get top performance.
> >>
> >> Which means my fstab contains the following line
> >>
> >> tmp     /tmp    tmpfs   size=10G,mode=0777      0 0
> >>
> >> and you'll notice the size=10G parameter, giving me a 10Gb /tmp directory.
> >>
> >> (I run gentoo, so /var/tmp/portage is also a tmpfs, and that's declared
> >> at 30Gb!)
> >>
> >> Cheers,
> >> Wol
> > 
> > In my /etc/fstab file I got this line on Slackware 14.1:
> > $ grep "tmp" /etc/fstab
> > tmpfs            /dev/shm         tmpfs       defaults         0   0
> > It is a bit different than your line in /etc/fstab but I don't know if that matters.
> > 
> > OK, so I see three things I can do in this order:
> > 1) Try to build libre office in an other directory than /tmp (because it is a tmpfs) and /home (because it is an NFS share)
> 
> With no size option, /tmp will default to half of ram. I would just add
> the "size" option, so change the fourth parameter to
> "size=10G,defaults", and you'll have a 10gig /tmp.
> 
> MAKE SURE that ram+swap is bigger, or a "/tmp is full" will crash your
> machine!
> 

It seems that my machine does not crash whereas currently:
 * I got this line in my /etc/fstab file:
tmpfs            /dev/shm         tmpfs       defaults         0   0
 * My RAM is 4GB
 * My swap is 1 GB
 * RAM + swap = 5GB
 * Calling '$ du -h /tmp' gives me a /tmp disk usage of 15GB (which is bigger than my RAM + swap)

Why my machine does not crash in these conditions?