Earmark.Global.Messages not started

[qcam]

My application uses Earmark in compile time and some times I get this error.

** (exit) exited in: GenServer.call(Earmark.Global.Messages, {:get, #Function<2.2630312/1 in Earmark.Global.Messages.get_all_messages/0>}, 5000)
    ** (EXIT) no process: the process is not alive or there's no process currently associated with the given name, possibly because its application isn't started

Have anyone encountered this before? Perhaps we need a way to supervise the agent?

[RobertDober]

Thank you for reporting this. Unfortunately I am only an OTP beginner, but the decision to use an agent for error messages has been taken because failure is acceptable (within certain limits of course).

My question is: Is the occasional failure a nuisance.

[ericmj]

The problem seems to be that earmark doesn't support being called from multiple processes concurrently. If you need to use a global, named agent I would suggest starting it in the application supervisor and storing error messages for each process instead of breaking when the agent is used from two processes concurrently.

[qcam]

@RobertDober @ericmj Thanks for your support.

Is the occasional failure a nuisance.

Personally it's a yes for me, how I occasionally got this error was calling Earmark.as_html/1 from multiple processes in compile time (I was not trying to access the agent manually).

I totally agree that failure should not fail the parsing, but in this case the error was from GenServer because when the agent was not found (when Earmark._as_html() tried to pop_all_messages() I believe).

The problem seems to be that earmark doesn't support being called from multiple processes concurrently.

I believe v1.1 does not have this problem. Sorry for my ignorance but maybe the introduction of
Earmark.Global.Messages wasn't a really nice improvement, since it took away the possibility of concurrent parsing?

And actually I think all texts are valid Markdown from the parser point of view? In that case maybe we don't need error agents?

[RobertDober]

@qcam I was not aware of this restriction, and if you can go back to 1.1 (or even 1.1.x) please do so.
But I will try to follow @ericmj 's advice (thanx for your ongoing support when my ignorance strikes BTW).

Well even in great programming languages like Elixir/Erlang, Clojure and Haskell, state is still an additional hassle. But I will definitely fix it, maybe getting pure again, not sure yet.

[qcam]

@RobertDober Thanks for the fix ❤️

[RobertDober]

@qcam can you live with the github version, or do you want @pragdave to make a release?

[qcam]

@RobertDober thanks. I think I'm good with 1.1.x so you can release whenever you wish. :)

[RobertDober]

We appreciate that.