How should we deal with invalid characters on 1.9?

[eric]

We ran into issues on 1.9 with a file that is supposed to be UTF-8 having invalid characters in it.

A fix was suggested for remote_syslog that should clearly go directly into eventmachine-tail, but I haven't been able to figure out exactly how I would want to fix it.

Here is the discussion that we've had so far: https://github.com/papertrail/remote_syslog/pull/13

Any thoughts would be welcome for the best way to solve this.

[jordansissel]

Since em-tail doesn't display or do any calculations on characters, really, I don't think it should care what encoding the data has that it is reading - so if it's breaking on some input, I think it's a bug in em-tail.

Can you publish a sample file with some bad data? Otherwise I'll try to reproduce and hack on a fix.

[eric]

I was just playing around with this: https://gist.github.com/1169737

[mblair]

I've hit this too. I've tried the iconv workaround in the remote_syslog pull request, as well as something like:

data = data.encode!( 'UTF-8', invalid: :replace, undef: :replace )

And I'm still getting the following error:

/usr/lib/ruby/gems/1.9.1/gems/eventmachine-0.12.10/lib/em/buftok.rb:66:in `split': invalid byte sequence in UTF-8 (ArgumentError)

Any ideas?

[vihai]

Any news about this issue?

[jordansissel]

Probably should just read into a buffer that is set explicitly to binary mode, and let the consumer of the data care about the encoding.

I'll get to fixing this eventually if nobody else does.

[rb2k]

Resurrecting this after a few years :) We just ran into this as well.

For us, we launched an app with start-stop-daemon and didn't pass the LC_ALL variable set to something UTF-8'ish

--> Ruby uses POSIX/ASCII and blows up when having to touch and UTF-8 char

[jordansissel]

This project has been replaced by the filewatch library. Last I knew, event machine was abandoned as a project (most recent release is 1.5 years ago), so I recommend not using em-tail.

sorry for the bugs, but this project is probably not worth resurrecting.

Recommend you check out the filewatch library instead, maybe?

On Thursday, September 4, 2014, Marc Seeger [email protected] wrote:

Resurrecting this after a few years :) We just ran into this as well

— Reply to this email directly or view it on GitHub https://github.com/jordansissel/eventmachine-tail/issues/13#issuecomment-54489680 .

[rb2k]

Sure, probably a good choice :)

Although I don't see an integrated way of actually tailing a file, rather than just being notified that something changed? But maybe it's just too early in the morning ;)