[#84] use admin flag to change AVUs when recalculating totals (4-2-stable)

[d-w-moore]

When plugin needs to modify AVU's that track quotas on behalf of a rodsuser (e.g. replica_close as part of istream write; see #84) some new logic will be required to make this work. Because setting metadata can mean redirection to the provider from the connected server, the definitive fix appears to involve the agent opening a client connection to self and this requires resolution of irods/irods#6782.

Same logic applies for running a general query as a rodsadmin; in this case, it was necessary to author a dedicated function to execute the query using the C API calls because irods::query is already included/linked but only in the server-enabled version.

[korydraughn]

That is correct.

[d-w-moore]

@korydraughn Tried the client conn for set_metadata but I'm getting this compile error:

$ make
Scanning dependencies of target irods_rule_engine_plugin-logical_quotas
[ 33%] Building CXX object CMakeFiles/irods_rule_engine_plugin-logical_quotas.dir/src/handler.cpp.o
/tmp/irods_rule_engine_plugin_logical_quotas/src/handler.cpp:599:21: error: no member named 'client' in namespace 'irods::experimental::filesystem'
                fs::client::set_metadata(fs::admin, clientComm, path, {attrs.total_number_of_data_objects(), objects.empty() ? "0" : objects});
                ~~~~^
1 error generated.
CMakeFiles/irods_rule_engine_plugin-logical_quotas.dir/build.make:75: recipe for target 'CMakeFiles/irods_rule_engine_plugin-logical_quotas.dir/src/handler.cpp.o' failed

[korydraughn]

What's left for this?

• Write tests
• Handle TODO comments (in code)
• Mark review comments as resolved

[d-w-moore]

Yes, looks like TODOs we added are done, will remove them.

[d-w-moore]

Test(s) yet to be written, starting with that now.

[korydraughn]

We need to apply similar changes to handlers which make calls to the following functions:

• get_collection_username
• switch_user

[d-w-moore]

@korydraughn Good point.

[d-w-moore]

One of those functions, unset_metadata_impl, calls get_monitored_collection_info, and that in turn has an irods::query using an rsComm_t&. Can we now do that as an rcComm_t& from a server plugin? You've adapted the filesystem calls along these lines. I wonder, do we need to do similarly now with the irods_query.hpp header?

[korydraughn]

We may have to do something similar for irods_query.hpp, but for now, let's use the C API directly for those cases where we need to make a GenQuery call using an RcComm. We'll deal with irods_query.hpp in 4.3.

[d-w-moore]

About to make an enhancement request in irods/irods (milestone should be >=4.3.1 ) relevant to the PR description above:

Same logic applies for running a general query as a rodsadmin; in this case, it was necessary to author a dedicated function to execute the query using the C API calls because irods::query is already included/linked but only in the server-enabled version.

[korydraughn]

Yes, >=4.3.1.

[korydraughn]

I see unresolved comments on this PR. Please mark them as resolved if you've handled them.

[d-w-moore]

done!

[korydraughn]

Is this ready for review?

[d-w-moore]

Is this ready for review?

Yes. Also: (1) Rebase done as discussed yesterday and (2) I've also run the tests on the cherry-pick to main, with everything passing in both branches.

[d-w-moore]

Tests still pass. All conversations appear resolved except for the one, will take care of that one today.

[korydraughn]

Do all tests pass?

[d-w-moore]

Do all tests pass?

All tests pass at the bench.

[korydraughn]

Good. Please squash to taste and we'll do one last pass over it.

[d-w-moore]

squashed. will squash in main later, after making the issue for the one failing test.

[d-w-moore]

# is added!