Fix grouping-by() use through parser references

[bazsi]

grouping-by() had an incomplete log_pipe_clone() implementation, which meant that whenever we were using a grouping-by() through a reference (e.g. not as an inline block), not all settings propagated to different references.

This branch is a much larger than I anticipated as it also implements state sharing between these clones and refactors a lot of similarities between db-parser() and grouping-by() as they share a common infrastructure.

I added a test to get at least some unit test coverage for grouping-by(), but that's still far from complete. db-parser() has a more complete testsuite and a lot of the code is shared between the code. I do hope that @kira-syslogng and light has some end-to-end tests.

Albeit this branch makes it possible it does not add similar state sharing to db-parser(), yet. I opened the PR as I felt I needed to get some feedback (from both reviewers and kira).

Also, it's a tricky question which of the clones would handle timer expirations. I do know the answer, but it's a great question to answer if you are doing the review.

I've used this config to test:

@version: 3.36


parser p_gby {
        grouping-by(
                key("$PROGRAM")
                aggregate(
                        value("aggr" "$(list-slice :-1 $(context-values $mod))")
                )
                timeout(3)
        );
};

log {
        source { tcp(port(2000)); };
        if ("$(% $SEC 2)" == "0") {
                rewrite { set("0" value("mod")); };
                parser(p_gby);
        } else {
                rewrite { set("1" value("mod")); };
                parser(p_gby);
        };
        filter { "$aggr" ne "" };
        destination { file("aggr.log" template("$(format-json aggr)\n")); };
};

[kira-syslogng]

Build FAILURE

[lgtm-com]

This pull request fixes 1 alert when merging 7ab37f3ba232312552d821709750bee1681db79f into ba656ba - view on LGTM.com

fixed alerts:

• 1 for FIXME comment

[kira-syslogng]

Build FAILURE

[bazsi]

IRL we discussed this topic and concluded that we shouldn't share state accross multiple references of the same grouping-by() instance.

[bazsi]

I have now removed the state sharing piece, as discussed, so this is ready for review/merge.

[kira-syslogng]

Build SUCCESS

[lgtm-com]

This pull request fixes 1 alert when merging 753b86a7afa1470f30b9aabc17a5e066dc049689 into 9ee50ad - view on LGTM.com

fixed alerts:

• 1 for FIXME comment

[kira-syslogng]

Build SUCCESS

[lgtm-com]

This pull request fixes 1 alert when merging ca197a5a177c08fa067e0361b86fdb72f464c835 into c5e823f - view on LGTM.com

fixed alerts:

• 1 for FIXME comment

[bazsi]

Rebased against current master. A review of this would be very much appreciated as I have some patches against grouping-by in parallel branches which have a high probability of conflicting with this one.

[kira-syslogng]

Build SUCCESS

[lgtm-com]

This pull request fixes 1 alert when merging 9b275471a89c8ae55874836a463533bd57c9d111 into 06fa9d7 - view on LGTM.com

fixed alerts:

• 1 for FIXME comment

[kira-syslogng]

Build SUCCESS

[lgtm-com]

This pull request fixes 1 alert when merging 8a610a3 into 8b7c37e - view on LGTM.com

fixed alerts:

• 1 for FIXME comment

[MrAnno]

Nice refactors.

We should write a news entry for this.

[bazsi]

added a news entry. thanks for the review.

[lgtm-com]

This pull request fixes 1 alert when merging 473a5d0 into 4aabae7 - view on LGTM.com

fixed alerts:

• 1 for FIXME comment

[kira-syslogng]

Build SUCCESS

[MrAnno]

Thank you.

[OverOrion]

LGTM, thank you!