…class

== context ==

Almost any RSolr exception is wrapped by Blacklight in a `Blacklight::Exceptions::InvalidRequest`

A Blacklight Catalog controller [rescues](https://github.com/projectblacklight/blacklight/blob/main/app/controllers/concerns/blacklight/catalog.rb#L21) this exception with a [method](https://github.com/projectblacklight/blacklight/blob/main/app/controllers/concerns/blacklight/catalog.rb#L282-L299) that just re-displays the search page, with the message "Sorry, I don't understand your search."

"Connection refused" errors are handled differently -- which is good, it would make no sense to prompt the user to search again with "Sorry, I don't understand your search" to network errors like connection refused.

What about a timeout error, where Solr did not return in a response within the timeout value RSolr had configured?  Prior to this PR, that was handled the same as any `Blacklight::Exceptions::InvalidRequest` -- redisplay search page to the user, with message "Sorry, I don't understand your search."

That is not a desirable UX, and also causes Solr timeout errors to generally not make it to many error monitoring and logging systems, since Blacklight rescues them and displays a 200.

== change ==

With the release of RSolr 2.4.0, RSolr uses a specific exception class to raise timeout errors, so Blacklight can rescue them and handle them differently.

After this PR, they are handled analogously to connection refused errors, with a custom exception that won't be rescued as a `Blacklight::Exceptions::InvalidRequest`.

a `defined?` check is done so Blacklight can still work with earlier versions of RSolr, and just won't do this new rescue.

== weird code in test ==

RSolr::Error api has some really rough edges, which requires some weirdness in setting up the test.

== to be done in future ==

I formatted the message of this exception just like the existing `Blacklight::Exceptions::ECONNREFUSED`... even though that formatting isn't necessarily so useful, there's a lot of stuff in there.

But I opted to do it consistently.

…tory agnosticism

… method

Thanks @barmintor, I think you're right this is more clear.

…cklight::Solr::Repository