Good morning from Montreal Awilliam
I
am retired after 55 years as system architect in ERP, banking, capacity
planning, and do hobby C programming and I also like to keep myself
active. I do enjoy doing some beta testing and "technical
writing/editing".
My
recent tests that prompted my email has its origins with the testing
of Fedora 27 betas and Gnome 3.26 I have been performing regular
re-installs and retests.
My
test method is to wipe clean a 100gig partition and perform an
installation. When I note repeated problems with a Fedora software, I
raise a bugzilla report. So, when I began to have problems with btrfs
and Gnome3.26, I redid the installation onto several other
disk/partitions. I then redid the installations using ext4, lvm or xfs
and additional btrfs installations.
Generally,
I install an extension and do invoke it's settings options. This is
what I noted. "As long as there is no extension setup via settings to
do, there is generally no lockup". I regularly test a Fedora beta or
Fedora nightly install for two to three days at a time before wiping it
clean and reinstalling a more recent version.
I have been trying to debug the extensions listed below by reading the code.
With
no extensions enabled, I generally experience zero problems at logon or
the rare random one that requires root's killall -u userid.
If
I enable Fedora delivered extensions, there is also usually no problem
(btrfs, ext4, xfs, lvm). I am not sure if that holds true if I
intentionally, under btrfs,recompile the schemas. I would, if you want
me to set up a test. Re testing, I have 5 terrabyte disks and ample
partitions available. I do not mix setups. This is what I tend to have.
sda=lvm, G3.26, 750gigs spare Fedora 27 beta
sdb data disk -- xfce F26 distribution thereon
sdc #1 G3.26 btrfs/btrfs (/ and /home), Fedora27 beta
sdc#2 G3.26 btrfs/xfs (/ and /home with xfs) Fedora27 beta
sdd G3.24 btrfs/btrfs (/ and /home) no problems Fedora 26
sde G3.24 ext4/ext4 (/ and /home) no problems Fedora 26
Note: Testing experience is causing me to make my conjecture about these added extensions.
All
the following extensions are tested with btrfs. xfs, ext4 and lvm.
Are four extensions are installed within the .local environment.
1)
Gnome Radio F27 Gnome3.26, with /home on btrfs -- Had to remove the
extension -- with it, no kbd, monitor, mouse. Needed a poweron reset.
2) TaskBar by zpydr. Would lock up Gnome, but I could get in via virtual terminal and do a cleanup.
3) gnomenu by Panacier (random installation/use )
4) Activities Configurator by nls1729 (random hiccup)
Are above extensions are installed within the .local/share environment.
What I have done to confirm for myself is: Without re-installing a full btrfs installed system, move the /home
to a partition formatted as ext4 and voila, all the faulty extensions
work. Via a two line change in /etc/fstab, I can toggle between the
btrfs partition and the ext4 other.
And the /usr/share/gnome-shell/extensions appear to suffer likewise.
The worst is do the following: glib-compile /usr/share/glib2.0/schema while with btrfs as the file system.
I regret I failed to do a diff with before/after of the compiled /usr/share/glib2.0/schema/.
With Fedora 26, gnome 3.24 btrfs, ext4, xfs, lvm, no matter the interface, all work with 3.24.
One
problem I am experiencing with Fedora 27 beta is the crash at login
time with gnome-shell going into a loop with 99.5% cpu. Many times I
power-off, restart, relog and experience a good logon and redo a crash
report. It seems often to require two logon attempts to be successful.
How do I trace execution? I would like to do the debugging.
Shortly I will be upgrading to Fedora 27, I will not be installing my own long term version with btrfs.
For
what it is worth. SUSE linux uses btrfs for / but xfs for /home. I
think that "that is the silent way they are not seeing extension
crashes".
By
the way, if I am wrong, for what ever justified reason, then I do
apologize for causing all this uproar. I am not intending to be a sxxx
disturber.
