Do not crash with internal error if /etc/hosts is corrupted #603
Conversation
|
Unit test would be more than nice to have... |
|
So the situation is that a config file contains garbage (which manifests as a parsing error down in CFA). It's not random bytes but YaST can't make sense of it. If we say "Reading /etc/hosts failed" the user will think that the file is not there or the permissions are wrong. We should clearly say that we do not understand the contents. |
There was a problem hiding this comment.
@jreidinger have you dealt with unparseable data yet? I would like to get your opinion to make this consistent.
src/modules/Lan.rb
Outdated
| @@ -370,7 +370,10 @@ def Read(cache) | |||
| # Progress step 6/9 | |||
| ProgressNextStage(_("Reading hostname and DNS configuration...")) if @gui | |||
| DNS.Read | |||
| Host.Read | |||
| if !Host.Read | |||
| ret = @gui && Popup.ContinueCancel(_("Reading /etc/hosts failed.\nThe file content is not in expected format.\nIf you continue any existing configuration will be overwritten.")) | |||
src/modules/Lan.rb
Outdated
| @@ -370,7 +370,10 @@ def Read(cache) | |||
| # Progress step 6/9 | |||
| ProgressNextStage(_("Reading hostname and DNS configuration...")) if @gui | |||
| DNS.Read | |||
| Host.Read | |||
| if !Host.Read | |||
There was a problem hiding this comment.
yast2 host will fail because it does not check the return value or Host.Read, and there are other callers too.
src/modules/Host.rb
Outdated
| # save hosts to check for changes later | ||
| @hosts_init = CFA::Hosts.new | ||
| @hosts_init.load | ||
| rescue RuntimeError => error |
There was a problem hiding this comment.
Two problems with this:
RuntimeErroris too generic- cannot see where the parse error occurred
In detail:
-
How about having an exception class across YaST,
GarbageDataError, meaning that YaST cannot understand the config files that it's supposed to edit. My point is that we should not let the user proceed and instead have a global rescue, similar to the (500) Internal Error, but labeled (400) "It's Your Fault!!", or "Lost In Translation", or "Cannot Process Such Data". -
The actual error is like
'Augeas parsing/serializing error: Iterated lens matched less than it should at /usr/share/augeas/lenses/dist/hosts.aug:23.12-.42:' (RuntimeError)
and a specific problem for Augeas is that the line location refers to the lens, or grammar, and not the actual unparseable data :( Line 23 in hosts.aug is
let lns = ( empty | comment | record ) *We probably cannot solve that in this PR. See also https://github.com/hercules-team/augeas/wiki/Tracking-down-errors
There was a problem hiding this comment.
ideally we shoud log content of file and then we can easily reproduce it.
|
@mvidner in bootloader I am solving it with this: https://github.com/yast/yast-bootloader/blob/master/src/lib/bootloader/config_dialog.rb#L23 |
|
The PR for reporting error location in the parsed file was already merged for Taking that in account I would like to know which is the state of this PR. Do we plan to continue working on it?, is the patch OK for SP3 as it is (of course after solving the conflicts)? The PBI for this PR is under the waterline but has not received much attention since a while: |
mchf
force-pushed
the
bnc1077435-internal-error
branch
from
5fc6792
to
32ccf01
Compare
|
backported updates from #621 |
mchf
force-pushed
the
bnc1077435-internal-error
branch
from
32ccf01
to
40add27
Compare
bsc#1077435