Bug #71372 for Net-ISC-DHCPd: Groups in dhcpd.conf

Hi Jan Henning, Am 03.10.2011 21:22, schrieb Jan Henning Thorsen via RT: Show quoted text
thank you very much for this quick response. Show quoted text
not as long as it is put out by "generate" as it was in the input. The fix is nearly perfect as it now generates all the missing lines. But there are still some problems: - The main problem that keeps me from being able to use your class is: the sequence of commands is mixed up in the output of "generate". This leads to the daemon not accepting the generated dhcpd.conf file. The subnet sections with the class definitions and the option statements defining the user options have to be in front of the host and host group definitions or the deamon will produce many error messages. - You don't output a trailing ';' after the '}' of the key section. Most examples show that but it doesn't seem to be required by the daemon. But very easy to fix ;-) ... - When specifying the keyname in double quotes the name contains the last double quote leading to two double quotes at the end in the output of generate which messes up the daemons parsing. That's perhaps because the regex matching when parsing the name in your key class is not greedy. This is a very minor issue as most examples don't use double quotes around the key name. But it is accepted by the DHCP daemon. - And I personally have a very special problem: I need comments to be retained (at least the "# monitor: 3 4 Y Studenten_DHCP" comment in the pool definition). That is because of the nagios plugin dhcp-pool ( http://folk.uio.no/trondham/software/dhcpd-pool.html ) needing these comments to work correctly. That's of minor priority as I can easily write a script that inserts those after generating the dhcpd.conf. I already worked around the problem of the disappearing comments by defining the user options which now contain the comments (I hope they have no side effects as I don't fully understand the concept of user options...). But that won't work with this plugin unless I change the source code (might be easy, but a bit of a hassle in case of updates of that plugin). Regards Michael

I have pushed a new release to CPAN now. Please check it out. 0.12 Tue Oct 4 15:06:11 CEST 2011 - [RT71372] Contributor: Michael Rath - Change child nodes has the same order as input - Config::Block has _body (ArrayRef) instead of body (Str) - Config::Block body text is indented as expected - Config::Key->_build_regex handle quotes - Add trailing ";" to Config::Key block (not sure why though...) - Add support for comments before elements. Generated output will also have comments See Net::ISC::DHCPd::Config::Role/comments for new attribute containing comments. t/40-rt71372-group.t now takes your input file and checks for the same output from generate() so I think it is OK now, even though I have made some changes to the input. The changes I made is adding whitespace in front of comments and config lines.

I had some spare time, so I thought that it was better to do it now, than later :-) My module is indeed limited, since it parses one line at a time, and the regular expressions used may be a bit too strict. That means that it may not be able to read "ugly" dhcpd.config files. Too help understand it's limits I add support for the ISC_DHCPD_TRACE environment, which will print what could NOT be parsed to STDERR. Very much like perl's warnings would do. Please set the variable to see what lines it fails on... (ISC_DHCPD_TRACE should probably be set to 1 by default?) The comment "Config::Block body text is indented as expected" is really, really bad. What I discovered is that the indentation was *better* after my latest commit, but i *knew* it was probably not perfect, since I'm removing the first four whitespaces... Patches are welcome :-) My goal was not to make really pretty config-files, but rather to be able to print something that was readable and parsable again by the module. Not sure if I will focus on this cosmetic change. Reason for that is I don't use ISC DHCPd myself, so I only fix bugs that are fatal because I feel that is the responsibility of module author... Comments at the end of block will be attached to the config line afterwards. The only comments which is not parsed, is the one at the end of file and "trailing" comments. Example: # this comment will be attached to the "host" object host foo { # this comment will be stripped away ... } # this comment will be attached to "option" object below option foo 42; # the last comment in the file will be ignored (Should probably add that to the documentation...) I do not want to retain all details. I just want to retain enough details. I think it is acceptable to require a format on the input text... So to sum up... Was there anything I needed to fix - except cosmetic bugs?

The reason was that Pool and Include had missing child elements. I just added the missing elements and created a new release to CPAN. Hope I haven't forgotten about anything... If you like, you're more than welcome to look over the list of childrens in each Config:: package to see if it makes sense. Look for: __PACKAGE__->create_children(qw/ ... /); (as you probably already know). 0.13 Wed Oct 5 12:43:42 CEST 2011 - Add custom warnings group for Role->parse() warnings. This can be silenced with (no warnings 'net_isc_dhcpd_config_parse';) - Add example config parser script - Add more child nodes to Config::Include * Net::ISC::DHCPd::Config::Key * Net::ISC::DHCPd::Config::Group * Net::ISC::DHCPd::Config::Block - Add more child nodes to Config::Pool * Net::ISC::DHCPd::Config::Host