Limit the size of the logfile

[happysalada]

Describe your feature request

I left the editor overnight somehow, I woke up to a 4GB logfile.
It seems something went wrong, but the logfile is filled with
2021-09-09T10:20:04.104 helix_lsp::transport [ERROR] err <-
the lines are truncated, so unfortunately I can't say much more.
How about limiting the size of the log file to 128 mb ? (just in case it gets out of control)

[kirawi]

Some alternative solutions may be to:

1. Not log empty errors.
2. Attempt to shutdown/kill the LSP process after a certain number of errors.
3. Have less being logged when logging is not turned on via -v.

[kirawi]

By the way, is this on master or 0.4.1? I believe it should have been fixed on master.

[sudormrfbin]

There's some discussion on the issue of log size in #566. Raising the log level to warn or error by default would also keep the size in check.

[happysalada]

it's on 0.4.1

[happysalada]

also perhaps not truncate the error (in terms of line size limit)? if it was a little longer, more information could be gathered I think.

[archseer]

There's no truncation happening, the error is just empty. If the LSP died we did not properly detect EOF and just kept reading 0 bytes and erroring. It's fixed on master 41f1e8e

Raising the log level to warn or error by default would also keep the size in check.

We actually do that already

helix/helix-term/src/main.rs

Line 12 in 0b1bc56

0 => base_config.level(log::LevelFilter::Warn),

[sudormrfbin]

Raising the log level to warn or error by default would also keep the size in check.

We actually do that already

Ah my log file was 20,000 lines long and there were INFO lines at the top from an older date, so I assumed they log level was unset.

There's no truncation happening, the error is just empty. If the LSP died we did not properly detect EOF and just kept reading 0 bytes and erroring. It's fixed on master 41f1e8e

Then I think this issue can be closed ?

[happysalada]

I'll re-open if this happens again with master.
Thank you for taking the time to take a look!

[smores56]

Unfortunately, I just ran out of space on my device from a 17Gb log file. I deleted it before I read it to check the contents, so I'll make sure to do that next time, but I have only been using Helix for a week or so. Having a limit on the length seems pretty valuable.

[kirawi]

I'm not sure if fern (the logging library we use) has built-in support for this, so we may have to roll our own logging struct.
daboross/fern#14

[smores56]

Yep, it was 2021-10-17T14:57:03.630 helix_lsp::transport [ERROR] err <- times a billion again.

[kirawi]

That specific bug was already fixed in master.

There's no truncation happening, the error is just empty. If the LSP died we did not properly detect EOF and just kept reading 0 bytes and erroring. It's fixed on master 41f1e8e

Raising the log level to warn or error by default would also keep the size in check.

We actually do that already

helix/helix-term/src/main.rs

Line 12 in 0b1bc56

0 => base_config.level(log::LevelFilter::Warn),

[smores56]

That specific bug was already fixed in master.

I installed from pacman, so I'll just compile from source this time.

[smores56]

Though this shouldn't be an issue for me, it still seems that having a non-truncating log file could be problematic here. fern doesn't currently have a way to auto-truncate, but it would be better IMO to add the functionality there rather than custom roll something here or change the entire logging library.

[kirawi]

It wouldn't be something that drastic if we do custom roll something.

[pickfire]

Probably we need to custom roll one ourselves, probably can use tracing-subscriber and write a custom MakeWriter. https://docs.rs/tracing-subscriber/0.2.25/tracing_subscriber/fmt/writer/trait.MakeWriter.html

[archseer]

Before we go overboard with the implementation, all we need to do is use https://doc.rust-lang.org/std/fs/struct.OpenOptions.html#method.truncate

[archseer]

log_file generates an `OpenOptions: https://docs.rs/fern/0.6.0/fern/fn.log_file.html

[kirawi]

It's not quite the same since it will completely wipe the log if you were to open helix again. This means that it doesn't prevent you from racking up a 17gb log in one session (though I guess it's only a problem if there was a bug like before). It's probably a good enough solution though.

[archseer]

#895