Race condition when running two elixir scripts in the same unix pipeline #13326
Comments
|
Good find. We need file system locking to solve this, which is not yet available in Erlang/OTP. So we will be able to solve this only in the future. |
|
I run into a similar issue in ElixirLS install script. In my case the race happens when multiple editor windows are started, each one running its own LSP process. I haven't solved that yet. As a side note piping into |
|
In your case @kidp330 you can workaround the issue with manual |
* Bump to 0.17.0 * Workaround for elixir-lang/elixir#13326
|
Hello, sometimes in developer mode I get this error in my iex and I can not debug this. it is returned automatically, and breaks my iex session I am using these are same error? @josevalim |
|
It seems to be something different. What is your OS and the first three lines of |
@josevalim I am very sorry, I cleaned my terminal, and I did not get this error again: My os: Intel macOs V14.3.1 but I remember this code: def create_user() do
data = fn ->
rand = randstring(10)
%{
"full_name" => "user #{generate(20) |> String.downcase()}",
"username" => generate(20),
"email" => "shahryar_#{rand}@gmail.com",
"scopes" => ["open_id", "email", "mobile"],
"devices" => ["ios", "mac"],
"status" => 4,
"roles" => [
%{"name" => "admin", "permissions" => ["admin:read", "blog:edit", "photo:read"]},
%{"name" => "user", "permissions" => ["blog:read", "photo:creat"]}
]
}
end
{time, _value} =
:timer.tc(fn ->
Enum.map(1..2500, fn x ->
case Users.create(data.()) do
{:ok, _struct} -> IO.inspect("User #{x} is created!!")
error -> IO.inspect(error, label: "User #{x} is not created!!")
end
end)
end)
time
endafter That I use this 2 or 3 times After 1 min it returned!! all process I did, returned success outputs. Thank you in advance |
|
@josevalim Finally I can figure out why it creates this error. When I write Persian in terminal and push the shift and a character after that delete it with Please see this Video 111.mp4Update in Elixir 1.17.0 (compiled with Erlang/OTP 27) |
|
This has been solved in v1.18 as it introduces a OS lock. You can see it here: |
Elixir and Erlang/OTP versions
Operating system
Debian 12 container
Current behavior
Scenario: I have two separate .exs scripts:
concat1.exs:
concat2.exs:
I would like to run both inside the same command, like so:
This works because they're using different caches, as you can see by the
Mix.install/2 using ...lines.Unfortunately the process of generating a hash to use for the cache is imo not strict enough, if I delete the
system_env:option from both scripts (or just set it to the same value) they will generate the same hash, and might fight over the cache. Example below:concat1.exs:
concat2.exs:
While the problem is less prevalent without
force: true, it may still happen if no cache yet exists:Expected behavior
The pipeline works as in the case where the scripts have unique
system_env's, (that is, separate cache directories).The text was updated successfully, but these errors were encountered: