Fix handling of trailing spaces in named.conf (bsc#976643) #54
Conversation
| @@ -22,7 +22,7 @@ | |||
| "comments" : [ "^[ \t]*#.*$", "^[ \t]*$" ], | |||
| "params" : [ | |||
| $[ | |||
| "match" : [ "([^ \t]+)[ \t]([^ \t]+(.*[^ \t]+)+)[ \t]*;$", "%s %s;" ], | |||
| "match" : [ "([^ \t]+)[ \t]([^ \t]+(.*[^ \t]+)+)[ \t]*;", "%s %s;" ], | |||
There was a problem hiding this comment.
what about using /spaces/*$ at the end?
There was a problem hiding this comment.
My first try was
[ \t]*;[ \t]*$
which also did the job. But then I found that one I used the even easier solution which the same outcome.
Honestly I am not sure if named.conf would even allow to have all the config in just one line without line breaks, but since our parser was not able to handle this even before this is at least no regression.
It is a L3 and so the priority should be a quick solution. The even better solution might be rewriting this using libaugeas – especially due to this comment in the code (by @jsrain):
// TODO: INI agent isn't optimal, in this file multiple occurences of one key
// in one section may be possible, section on one line is handled as a
// single value, and many many other problems are to be expected....
// should be replaced in future by some better agent
But this might better be a separate PBI.
There was a problem hiding this comment.
I tried a few scenarios and the solution that even would cope with comments after the semicolon would be:
[ \t];.*$
Both of the above versions fail on that and also the current implementation.
So with this one I would even fix a not-yet-found bug - although the parser needs improvement anyway.
|
The ini-agent parser is not very suitable for the config file. But given the bug, I think this limited fix is a good one. |
I tested that fix manually in a SP1 VM using the named.conf attached to the bug: It did not merge those two lines as it did before. The file looks sane afterwards.