Each subsystem needs it's own configurable timeout

[devinus]

For example, the code server doesn't need the same timeout as the compiler, etc.

[alexrp]

The compiler has no timeout. That timeout is used for the code server.

[alexrp]

Is there any actual use case for a separate timeout for each subsystem, anyway? Unless there's a good case for it, it just seems like overengineering to me.

[devinus]

People just have different needs.

[josevalim]

I have reverted #778 until we agree on the proper solution for this.

Indeed, I did not feel comfortable with having one shared timeout, I would also prefer if Mix and ExUnit have their own timeouts. Again, the simplest approach in my opinion is to have an arbitrary large timeout (30 seconds) value for those, without a configuration option. YMMV though.

[yrashk]

HUGE 👎 for reverting #778. It is a fine internal implementation detail
that can be incrementally improved later on.

On Thu, Jan 24, 2013 at 12:48 PM, José Valim notifications@github.comwrote:

I have reverted #778 https://github.com/elixir-lang/elixir/issues/778until we agree on the proper solution for this.

Indeed, I did not feel comfortable with having one shared timeout, I would
also prefer if Mix and ExUnit have their own timeouts. Again, the simplest
approach in my opinion is to have an arbitrary large timeout (30 seconds)
value for those, without a configuration option. YMMV though.

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/791#issuecomment-12672533.

[devinus]

I think the general idea of #778 is fine, it just needs to be per subsystem. Also, exposing System.services_timeout seems bad, am I the only one? It's just so generic, and I hate bloating the System module more just because we can't figure out better places to put things.

[devinus]

I think we need an Env module.

[josevalim]

If we are breaking into apps, there is no need for System.services_timeout.
After all, it is fine for elixir internal functions to access
:elixir_compiler or :elixir_code_server to get this info.

[alexrp]

I won't have time to implement a different approach to setting timeouts for the next couple of weeks (due to moving and picking up a new job) so unless someone else is going to work on this, please re-merge that pull request. I have no problem with going with a different approach but reverting the pull request with no one having any obvious plans to work on an alternative is not a sensible way to go forward with this IMO.

(Effectively my builds are back to square one with timeouts now...)

Yes, I know it means exposing a function in 0.8.0 that will go away soon after, but it's undocumented so I think it isn't that big of an issue to just change/remove later when I get some time. Or, as said, if someone else is going to work on this, please do.

[josevalim]

The reason for revert is not System.timeout_services but the exposed environment variable that may no longer be respected as people expect. I am going to provide a small patch that sets a higher timeout (30 seconds). We can discuss configuration options later.

[josevalim]

Just pushed ba528f0 that adds a default timeout of 30 seconds.

[josevalim]

I am closing this. If there is a need for configurable timeouts beyond 30 seconds, let's reopen this discussion.