[OpenIndiana-discuss] MATE 1.20 and updated GTK+3 for hipster

Tue Oct 23 14:05:14 UTC 2018

On Tue, Oct 23, 2018 at 3:45 PM Predrag Zečević - Technical Support Analyst
<predrag.zecevic at 2e-systems.com> wrote:

> On 10/23/18 15:02, Udo Grabowski (IMK) wrote:
> > On 18/10/2018 16:59, Alan Coopersmith wrote:
> >> On 10/18/18 05:55 AM, Udo Grabowski (IMK) wrote:
> >>> This is strange, why in the world should the settings-daemon
> >>> read the fontpathes in an infinite loop ? Looks like its constantly
> >>> thinking the mate font configuration has changed.
> >>
> >> I don't know about infinite loop, but you'll see libfontconfig iterating
> >> through all the fonts in all the directories when it thinks its font
> >> caches are out of date.   On Solarish systems, we have an SMF service
> >> svc:/application/font/fc-cache to ensure the caches are up-to-date at
> >> system boot and when IPS packages with fonts are installed, so it would
> >> be good to check that's running and not reporting any errors in the SMF
> >> logs for it.
> >>
> >
> > For what I see from the 2483 time-sliderd truss output, it does the
> > following for each /etc/X11/fontpath.d/*/ directory:
> >
> > 1) with a convoluted procedure, produce a lock file .uuid inside
> >    that directory
> >
> > 2) Then use utimensat to update access and modification times of
> >    that directory   *********
> >
> > 3) Open RDONLY  *.cache-n file in /var/cache/fontconfig/
> >             /export/home/predrag_zecevic/.cache/fontconfig/
> >             /export/home/predrag_zecevic/.fontconfig/
> >    (there are none)
> >
> > 4) Do the same RDWR (there are still none)
> >
> > 5) fstat all .pcf files and fonts.dir and files mentioned therein
> >    in the fontpath.d dir and mmap them
> >
> > 6) do something (not seen in truss) and unmap them
> >
> > 7) remove the .uuid file
> >
> > 8) repeat from 1)
> >
> > Could it be that the change of the modification time (2) from 2483
> > triggers the reread by 2484 (mate-settings-daemon) ?
> >
> Hi,
>
> many thanks fo looking at truss files!
>
> No idea modification time changes triggers more actions (could it be)...
> New MATE has (obviously) introduced more checks/processes/etc.
>
> Since, this is narrowed to fonts re-reading/re-caching - may I ask you
> to share contents (at least figures) from your desktop, of font
> directories (and packages)?
>
> > And by the way, if I counted correctly, this system has over 24.000
> > individual fonts, so the caching procedure a) really takes a while,
> > and b) makes no sense at all since almost all of them are obviously
> > unused. And indeed, the 2484 truss shows no loop up to the truncation
> > point, these are all different fonts...
>
> No idea, to me, this is not so big number:
>
> $ for dir in /var/cache/fontconfig
> /export/home/predrag_zecevic/.cache/fontconfig /etc/X11/fontpath.d; do
> printf "%50s:" $dir; find ${dir} -type f | wc -l; done
>                              /var/cache/fontconfig:408
>     /export/home/predrag_zecevic/.cache/fontconfig:199
>                                /etc/X11/fontpath.d:0
>
>
> [Latest one contains just links to some fonts, presumably created by
> fc-cache service]. Here count of sym-links:
> $ for dir in /var/cache/fontconfig
> /export/home/predrag_zecevic/.cache/fontconfig /etc/X11/fontpath.d; do
> printf "%50s:" $dir; find ${dir} -type l | wc -l; done
>                              /var/cache/fontconfig:0
>     /export/home/predrag_zecevic/.cache/fontconfig:0
>                                /etc/X11/fontpath.d:90
>
>
> Creation time of caches:
> $ ls -l /var/cache/fontconfig | grep -v ^total | awk '{printf("%3s %2s
> %4s\n", $6,$7, $8)}' | sort | uniq -c | sort -rn
>     197 Sep 21 11:46
>     196 Sep 21 11:56
>       9 Sep 21 11:49
>       3 Sep 21 11:29
>       1 Sep 24 09:04
>       1 Oct 23 13:39
>       1 Oct 23 13:37
>
> $ ls -l /export/home/predrag_zecevic/.cache/fontconfig | grep -v ^total
> | awk '{printf("%3s %2s %4s\n", $6,$7, $8)}' | sort | uniq -c | sort -rn
>     188 Jan 11 2016
>       8 Mar 14 2016
>       2 Feb 10 2016
>       1 Jan 19 2016
> [NOTE: /export/home is not located within rpool, so this stays same
> across BEs]
>
> *BUT*, if I mount problematic BE:
>
> $ pfexec beadm mount oi_181018 /beadm
>
> and count files:
> $ find /beadm/var/cache/fontconfig/ -type f | wc -l
> 121138
> [I can imagine that, if I have left system running long enough, this
> will be much bigger number]
>
> $ find /beadm/etc/X11/fontpath.d/ -type l | wc -l
> 90
>
> So, font cache fails to "recognize" fonts, and just keeps system busy
> with creating cache, over and over.
>
> How to solve that situation?
>
> With best regards.
> Predrag Zečević
>
> P.S: (/beadm is mount point of BE which behaves slow)
>
> $ du -shc /beadm/var/cache/fontconfig/ /var/cache/fontconfig/
> 152M    /beadm/var/cache/fontconfig/
> 15M     /var/cache/fontconfig/
>

Would the last two commits applyto your case?
https://gitlab.freedesktop.org/fontconfig/fontconfig/commits/master

> --
> Predrag Zečević
> Technical Support Analyst
> 2e Systems GmbH
>
> tel: +49 - 6196 - 95058 - 15
> mob: +49 - 174 - 3109288
> fax: +49 - 6196 - 95058 - 94
> e-mail: predrag.zecevic at 2e-systems.com
>
> headquarter: 2e Systems GmbH, Koenigsteiner Str. 87, 65812 Bad Soden am
> Taunus, Germany
> registration: Amtsgericht Koenigstein (Germany), HRB 7303
> managing director: Phil Douglas
>
> http://www.2e-systems.com/ - Making your business fly!
>
> _______________________________________________
> openindiana-discuss mailing list
> openindiana-discuss at openindiana.org
> https://openindiana.org/mailman/listinfo/openindiana-discuss
>

-- 
---
Praise the Caffeine embeddings