rustlings watch fails silently

[jounathaen]

Hi,
I can't get rustlings to work correctly. If I run rustlings watch, the program fails silently.
Apparently the problem is, that

rustlings/src/main.rs

Line 174 in 523d18b

watcher.watch(Path::new("./exercises"), RecursiveMode::Recursive)?;

fails on my system (Fedora 32, nothing special AFAIK) and this error only handled with an is_ok() in

rustlings/src/main.rs

Line 122 in 523d18b

if matches.subcommand_matches("watch").is_some() && watch(&exercises, verbose).is_ok() {

which then fails silently.

The error of the failing watcher.watch(/*...*/) is Io(Os { code: 28, kind: Other, message: "No space left on device" })' which is most likely a bug in notify as there are hundreds of GB free on the disk. However, I think the error message should at least be propagated and printed to the command line.

Cheers!

[jounathaen]

The problem lies apparently in the inotify watch limit. All of them were occupied (by my languageclient). So this is not a bug of rustlings. Only the lack of an error message is

[victordeleau]

imo I think the auto refresh should be replaced with a manual refresh (press enter refreshes by default for example), and that for 3 reasons:
1- People (like me) might get the issue described here (which I solved by increasing the inotify watch limit of my OS). But we can't expect people to tweak their OS like that to be able to use rustlings.
2- It is bothering when the rustlings watch does its thing without us having a say. Sometimes it is not exactly clear if the error displayed is related to the current exercise or the next one if some modifications made the tests pass. We might also want to read the error again even though we edited the code. That's especially true if the shell running rustlings watch is not in sight of the editor.
3- Pressing enter to refresh is a no-brainer that takes no time.

[seeplusplus]

Looks like this issue may have a resolution soon.

notify-rs/notify#266

[jyn514]

@seeplusplus that issue is just about the error message, there will still be an error either way. The issue here is that rustlings is ignoring the error, which notify can't do anything about.

[jounathaen]

By the way, if somebody want's to fix this issue. It is probably very easy. The code just has to be changed to something like:

if matches.subcommand_matches("watch").is_some() {
     if let Err(e) = watch(&exercises, verbose) {
        println!("Error: Could not watch your progess. Error message was {:?}", e);
        println!("Most likely you run out of disk space or you 'inotify limit' is reached");
        std::process::exit(1);
    }

I just don't have time to test this properly...

[seeplusplus]

@seeplusplus that issue is just about the error message, there will still be an error either way. The issue here is that rustlings is ignoring the error, which notify can't do anything about.

True. I guess I mean it's solved inasmuch as it responds with a "more correct" error.

I'll try my hand at adding @jounathaen's code and testing it to see if we can resolve this issue.

[seeplusplus]

I just don't have time to test this properly...

For the sake of correctness, what kind of testing would you recommend? I'm not super familiar with testing in Rust (is there some integration test that could be written to test this?). I did test that this gives us the behavior that we want (I dropped the max_user_watches to 256 on my machine and ran "rustlings watch").

[jounathaen]

That's exactly what I imagined as test.