Is it really the "right thing" to have erlang:halt(0) inside project@@main:run(project) ?

[oderwat]

I just wanted to use gleam.shell() to

eprof:start().
eprof:profile(fun() -> project@@main:run(project) end).
eprof:analyze().

But because of project@@main:run(project) uses erlang:halt(0) which shuts down the node, this is not possible.

I also thought it would be easy to use the existing run_module() for running multiple Gleam projects on the same node. But having erlang:halt(0) makes it impossible to use the prepared Erlang code for that.

Maybe the code should have two points to run it, where one is doing the halt and the other just returns?

[GearsDatapacks]

Can you not call project:main here?

[oderwat]

Can you not call project:main here?

Not really, the run() and run_module() code is configuring the VM for utf8 handling and makes sure the application dependencies are loaded. It also runs it as an isolated process. This can all be done "manually" from the shell too. As can the profile be run inside the Gleam code (which is what I implemented for this by now). But having a slightly different setup in the compiled code one could make it easy to do some stuff in the shell or use that code for deploying and running it in existing nodes.

[lpil]

project@@main:run/1 is not a public API, you must not use it directly and it is not guaranteed to exist.

[oderwat]

I see. I originally got it as part of Erlang shipment as part of the entry point scripts. I still think it would be beneficial to rearrange this code a bit, so you can run it without halting the node.

[lpil]

Again, this is not a public API so there's reason to change it in any way for any unintended use.