Compiled modules have the wrong bytecode filename

[ericmj]

:code.which/1 returns the name of the source file instead of the produced beam file.

[wkhere]

Yup I ran into the same issue when coding my Reprise module reloader.

A note which may help sorting out various cases - when you explicitly load a module via l iex shell command, the module gets loaded again with a proper filename pointing to the beam.

[wkhere]

Works in my case w/ Reprise. Thx!

Just curious what are :in_memory modules - isn't that every module must've been compiled somehow? :)

[josevalim]

:in_memory is for when the module exists only in memory and the beam file cannot be located elsewhere.

[josevalim]

Btw, thanks for reporting back that it works @herenowcoder !

[wkhere]

Now that 0.15.1 is released (kudos!) I experimented with it a little bit more, again in order to have good defaults for Reprise reloader.

What I see is when a module is first used in iex it has the :in_memory mark. The beam file shows up in a result of :code.which only if I load it via l.
Now I'm not sure if it changed between master commit I tested before and 0.15.1, or not. Anyway @josevalim could you drop a few words on the idea behind :in_memory - does it reflect some Erlang behaviour underneath or is it a way Elixir handles module loading?
TIA

[ericmj]

How is the module that has the :in_memory mark defined/loaded?

[wkhere]

Huh... Gentlemen, sorry for taking your time. Looks like I forgot to mix clean ;P. So no unexpected :in_memory, actually...

[wkhere]

..still I found one rare case:

Module is a part of supervision tree in OTP app so it should be loaded right away.

When I run mix clean; mix; iex -S mix the beam file is returned by :code.which.

Upon mix clean; iex -S mix the code is compiled, the app is running inside iex, but :code.which returns :in_memory.

Same goes for the modules which are not supervised, but are loaded into iex via function call. If the code was compiled only when iex -S mix was starting, there's :in_memory returned.

I can live with this; reporting for you to decide if this is what you intended.

[ericmj]

I can reproduce. That does seem like a bug. /cc @josevalim