Test: SIGINT should be able to interrupt non-terminating goals on linux

[danilp-id]

Test for #3290.

Currently fails, but passes if that PR is applied.

Works by creating a child scryer process wrapped in script to simulate a terminal and then sending INT signals to it. (If terminal is not simulated, bug will not appear.)

Also reworked tests to use #[tokio::test(flavor = "multi_thread")] instead of custom tokio initialization code.

TODO (in a different PR): currently, there is no way to detect if test encounters a deadlock. We need to decide how to test for timeouts as cargo test doesn't seem to support this out of the box.

TODO: this test uses script which is part of util-linux and thus not available on MacOS. Need to restrict this test just to linux or find a way to simulate terminal on MacOS.

[danilp-id]

d307981 hangs up the CI https://github.com/mthom/scryer-prolog/actions/runs/24620777859

[danilp-id]

See https://unix.stackexchange.com/questions/249723/how-to-trick-a-command-into-thinking-its-output-is-going-to-a-terminal for ways to simulate terminal. It also includes a way to make it work on all unixes by injecting a dummy shared library.

[danilp-id]

This might work for all unixes and only require python as dependency https://unix.stackexchange.com/questions/249723/how-to-trick-a-command-into-thinking-its-output-is-going-to-a-terminal#answer-564841

[danilp-id]

Huh, that's really weird. Why does it now pass on everything except for macos and ubuntu x86? It still fails if I run the test locally. Can someone check?

[triska]

I don't know why it fails on these systems, but one general comment: The Python script looks so small that maybe it could become a Scryer program? Is there anything we need for example in librayr(os) so that Scryer can express this?

[danilp-id]

@triska we can do it like exec:run/2 from https://hexdocs.pm/erlexec/exec.html#t:cmd_option/0 and add a pty option for process_create to wrap launched process with a pseudoterminal.

[Skgland]

@triska we can do it like exec:run/2 from https://hexdocs.pm/erlexec/exec.html#t:cmd_option/0 and add a pty option for process_create to wrap launched process with a pseudoterminal.

I don't think it's currently possible to configure that using the underlying rust API.

[danilp-id]

Yes, but we can use something like portable_pty. As a bonus, it also supports windows.

[abmclin]

@danilp-id thanks for this test case! It's great to be testing Scryer's REPL behavior across multiple platforms. It would be beneficial to have a complete test suite for as much of REPL's user interaction as possible to detect future regressions like the one caused by updating rustyline I had trouble adding test cases for Windows tty issues so it'll be interesting for me to see if the approach you used translates well to Windows.

[danilp-id]

@abmclin Thanks! The rustyline issue is unix-only, but on windows there is ConPTY which I think can be tested using pywinpty for example. I think for now it's ok to just make a small python wrapper script to test issues like these instead of adding pseudoterminal support directly to process library (and I am not sure if it would really be that useful to add it there since anyone can just use a wrapper script if the need arises).