WIP: Support more bind9 utilities #1321
Conversation
host and nslookup are already supported. But dig is not, similarly with delv and others.
|
https://gitlab.isc.org/isc-projects/bind9/-/issues/4879 @yedayak seems to have already shown interest in implementing it in the upstream project, and the upstream maintainer seems to be interested in having and maintaining the completions in their repository. If @yedayak has already started working on them, those two versions could be combined. After that, we can submit it to the upstream project. |
completions/bind9
Outdated
|
|
||
| case $prev in | ||
| -c) | ||
| _comp_cmd_nslookup__queryclass |
There was a problem hiding this comment.
This function is defined in another file, which may not be loaded in the session. One needs to "export" the function by renaming _comp_cmd_nslookup__queryclass to _comp_xfunc_nslookup_compgen_queryclass and call it using _comp_compgen -x nslookup queryclass.
This means that we need to modify bash-completion too even if we'd include the completion settings to the upstream.
There was a problem hiding this comment.
well, yes. unless we would rely on named-rrchecker providing list of supported types. That does not come with common package containing utilities only.
There was a problem hiding this comment.
further more, it might be useful also for ldns drill tool or unbound-host. Even though better would be getting actually supported list from the tools itself, we do not have it always at hand.
But the question is, whether it would not make sense to move nslookup functions into the bind9 as a whole. They are after all bind9 utilities in normal circumstances.
There was a problem hiding this comment.
But the question is, whether it would not make sense to move nslookup functions into the bind9 as a whole. They are after all bind9 utilities in normal circumstances.
We haven't been officially doing that when there are multiple implementations for a specific command name, but I think it would make sense. We should still keep our current implementation of completions/nslookup at completions/_nslookup for the other implementations of nslookup. We've been regarded our completion/_<cmd> completion settings as "deprecated" completion which should be replaced by the newer implementation of the upstream completion settings, but the present suggestion is to use completion/_<cmd> as the "default" completion settings when a specific completion setting is not provided by the upstream project. Or we might add even another form of the completion files for the "default" completion or another directory for the "default" completion. Thus, your suggestion involves a reconsideration of our framework about the role of our completion/_<cmd> settings or introduction of the mechanism of defining the "default" completion.
Partial fixes from review.
Adds multiple types specification for dig. It suggests qtypes and query classes (only uppercase!), known hostnames in addition. Add also quite similar support for knot's kdig, which is more or less directl dig alternative.
I have a start, but it mostly just has sed commands to extract |
Except kdig, they have different parameters presentation. Use simple awk to print parameters offered.
Use more hints from PR, use existing help with specified parameter.
Selected local variables from caller functions are accessible in nested functions, no need to pass them inside. Pass just remaining parameters.
host and nslookup are already supported. But dig is not, similarly with delv and others.