Roger Lipscombe

> 'proc_lib' or 'sys' might be a good place for this functionality with set/get_process_description/1. Is there a compelling reason to choose one over the other? Based on a quick skim...

Just to be clear: I've not forgotten about this; been busy with the day job.

I've updated the PR. See the new commit.

I've got time to work on this now, but I'm waiting for some feedback on the implementation.

If we discard the idea of `sys:get_label`, etc., and lean into using the process dictionary, can we make only one copy of the dictionary? I prefer using a well-known value...

> But maybe move the functions to `proc_lib`. Can do, but... > add a new function there, that lookup the label and if it does not exist lookup the initial...

Well, actually, it turns out, I _do_ get an erl_crash.dump file, but it's truncated.

I can reproduce this: ``` $ kerl build git https://github.com/erlang/otp.git OTP-20.3.2 OTP-20.3.2 Checking out Erlang/OTP git repository from https://github.com/erlang/otp.git... Building Erlang/OTP OTP-20.3.2 from git, please wait... Erlang/OTP OTP-20.3.2 from git...

Note that my experiment uses `realpath` from `/usr/bin`. The `kerl` script defines its own `realpath` function. I don't know if they're exactly the same, but if they're similar in this...

fwiw, I got file upload working with the following: ``` FName = hackney_bstr:to_binary(filename:basename(Path)), Disposition = {, [{name, }, {filename, FName}]}, ReqBody = {multipart, [ {file, hackney_bstr:to_binary(Path), Disposition, []} ]}, {ok,...