Add support for FreeBSD #48
Conversation
Using Hiera hierarchy, we can provide a default value that makes sense for the current operating system and still allow the user to provide their own custom package name if necessary. Makes the code simpler without changing the logic.
This directory is operating-system dependant.
These directory only make sense on some operating systems and should already be part of base system or the syslog-ng package. Do need to manage these in the patterndb module.
According to hier(7), /var/cache/ is used to store miscellaneous cached files, precisely what merged patterns not yet installed are.
|
Looks like CentOS images are gone. Time to remove support for this EOL system. I will open a PR to do so on top of this one. |
There was a problem hiding this comment.
Thanks @smortex I understand you are improving the semantics of the variable name here. While this is fine by me, it will require a major version bump.
Question: why do you change the smoke tests' base_dir too ? The reason I did this in the first place is so I could run puppet unprivileged for smoke testing.
Ah! I failed to understand this purpose. The different default value and value used in the smoke tests, and the weird value in the test suite just confused me :-) Now, I see your point, when there was no CI, it was indeed necessary. Now the integration tests make our life easier, so it makes sense to remove it completely.
Correct, I removed a parameter, so will change the PR accordingly. I just spot that I forgot to remove the default value from Hiera so will add a commit to finish this properly. |
The variable was removed, so no need for a default value and we should not see references to it in the README.
|
The README is still not exactly up-to-date. puppet-string will help to keep things in sync, but it is a big task and this PR is already big, so I suggest to switching to it when this PR is merged. |
There was a problem hiding this comment.
while I agree the base_dir isn't needed for CI, there are people still left in this world (guess who) who like to use actual bare metal containerless machines to actually run some software.
Why not name it _base_dir to stress it's undocumented/unsupported but still leave it around for those rare nerds who like to do it "the old way" ? 👼
| ``` | ||
|
|
||
| This *File* is generated by merging all defined *ruleset resources*, which come in two forms: [raw](#defined-type-patterndbrawruleset) and [simple](#defined-type-patterndbsimpleruleset). | ||
| Merging is handled under the hood by using `pdbtool merge` which creates a new *patterndb parser* in the `${temp_dir}` directory. Testing of the merged *parser* is optionally handled using `pdbtool test`. If this is a success, the merged file is then being deployed into the definitive destination at `${base_dir}/var/lib/syslog-ng/patterndb/${name}.xml`. | ||
| Merging is handled under the hood by using `pdbtool merge` which creates a new *patterndb parser* in the `${temp_dir}` directory. Testing of the merged *parser* is optionally handled using `pdbtool test`. If this is a success, the merged file is then being deployed into the definitive destination at `/var/lib/syslog-ng/patterndb/${name}.xml`. |
Of course I see this message only after merging the approved PR that include it 😑 I'll open a PR to revert it. |
Question: rather than prepending $base_dir = '/tmp/x'
class { 'patterndb':
cache_dir => "${base_dir}/var/cache/syslog-ng",
var_dir => "${base_dir}/var",
config_dir => "${base_dir}/etc/patterndb.d",
}That still* does not work out-of-the-box because |
|
yeah that would be all right for my use-case. |
More local patches I have been using on my FreeBSD systems and are worth merging upstream.