Register a listener(reporter) that always listens

[claremacrae]

Description

Currently, registering a new reporter only works if the program's command-line arguments request the new reporter.

This is great for a reporter, but it doesn't allow for a subscription model.

For ApprovalTests.cpp, we are looking to say REGISTER_LISTENER(MyReporter) - that will always listen, and not interfere with any use of --reporter on the command line.

Steps to reproduce

If you call REGISTER_REPORTER() without the use of --reporter=, your listener is not used.

Extra information

We (@isidore and I) would love to have a 10 minute Skype call with you, if you are interested.

We are doing this as someone has asked us to make ApprovalTests.cpp support doctest: approvals/ApprovalTests.cpp#11

[claremacrae]

On reflection, having looked at the Catch2 code more closely, I see that Catch has two separate concepts, Listener and Reporter.

Catch2 listeners are registered with CATCH_REGISTER_LISTENER() - and are always active. That's what we are using with ApprovalTests.cpp's Catch integration.

Catch2 reporters are selectable by the user at run-time, depending on their needs...

If I understand correctly, doctest has combined these two features in to one - hence this question.

[onqtam]

I think it should be doable like this:

context.addFilter("reporters", "my-listener");
context.addFilter("reporters", "console");
// and then call
context.applyCommandLine(argc, argv);

when providing a custom main.

But I'm not sure if that way the console reporter will be replaceable - perhaps the API isn't sufficient yet. Thanks for reporting this - I'll think about it some more when I have the time.

[onqtam]

I just pushed this in the dev branch - just use REGISTER_LISTENER instead of REGISTER_REPORTER - the behavior will be the same as in Catch2: all listeners will be executed before any reporter, and cannot be filtered through the command line.

I also just noticed your note about a skype call - I contacted your colleague on Linkedin about a google hangouts which we will arrange there I guess :)

[claremacrae]

Thanks! I read the differences in the docs, and I understand the changes because of this request...

I'd be happy to make some minor clarifications to the new docs if you'd be interested - what would be the most convenient way for me to do this?

Comments on the commit?
Fork the repo and make a PR on the dev branch?
Anything else?

[claremacrae]

Or pair?

[onqtam]

Sorry I left the office. Whichever is easiest for you - comments on the review work too :)

[claremacrae]

I can confirm that when I build ApprovalTests.cpp against the doctest single header on its dev branch, and change REGISTER_REPORTER to REGISTER_LISTENER, I can then switch to using the following, and can delete our custom main altogether.

#define DOCTEST_CONFIG_IMPLEMENT_WITH_MAIN
#include <doctest.hpp>

This is tremendous - thanks very much indeed.

I know you're busy, but do you have any idea when you might do your next release, please?

It would be lovely to get this included in ApprovalTests.cpp before I head off for CppCon, to plug the doctest integration as part of my talk there... :-)

[claremacrae]

I've also confirmed that the existing reporter mechanism works fine when I have registered a Listener as described above.

For example, running our tests with --reporters=console,xml does do invoke our listener, and does also write out the output of the console and xml reporters. :-)

[claremacrae]

As far as I'm concerned, this feature is good to go, and the issue could be closed - thank you again for implementing this! /CC @isidore

[onqtam]

I'll release an official version hopefully within 1-2 weeks - way before CppCon :)

[claremacrae]

That’s ideal - thank you!