GUACAMOLE-???: Terminal's OSC parser to not stop on unhandled characters

[krobelus]

** I haven't made a ticket yet because I'm still waiting for my JIRA account to be created ** untested; maybe someone else can test this by running the printf command below?

As reported in https://github.com/fish-shell/fish-shell/issues/11749, Guacamole's terminal fails to parse some OSC sequences. For example, running this command (in a shell like Bash):

printf '\033]133;A;special_key=1\007'

echoes part of the sequence which is incorrect.

Almost every other terminal ignores such sequences, even if they don't handle them. This allows apps to add new features without probing for support (which is not always possible or reliable).

This behavior was introduced in c56412f6 (GUAC-734: Return to echo after HTS. Return to echo on unexpected characters within OSC., 2014-06-11) for unknown reasons -- I don't know how to find that ticket (it's not GUACAMOLE-734).

[necouchman]

This behavior was introduced in c56412f (GUAC-734: Return to echo after HTS. Return to echo on unexpected characters within OSC., 2014-06-11) for unknown reasons -- I don't know how to find that ticket (it's not GUACAMOLE-734).

That ticket - GUAC-734 - comes from the old Glyptodon Jira issue tracker, before the project was part Apache community, so you won't find the ticket in the current Jira tracker.

The change looks simple enough to me; however, I wonder what the implications of it may be in other situations. Right now it just echoes the character(s) out to the terminal. I'm not sure transparently ignoring them is 100% the right way to go.

I'm thinking that:

• At the very least, the unknown characters should still be logged at at the DEBUG level (or perhaps the WARN level, if we're changing it to not even print the characters?).
• Maybe there should be a new configuration parameter for the protocols that use the terminal that allows toggling whether or not the characters are ignored? Perhaps a checkbox that says "Do not ignore unknown characters", where the default (unchecked) behavior is what you're proposing, here, that the characters are ignored, but one could tick the box and have it print those characters to the screen?

@mike-jumper, thoughts?

[krobelus]

That ticket - GUAC-734 - comes from the old Glyptodon Jira issue tracker

I wonder if we can find it; just in case there was a particular reason this was introduced -- though I don't expect there is, probably it was an insignificant change due to a misunderstanding.

I'm thinking that:

• At the very least, the unknown characters should still be logged at at the DEBUG level (or perhaps the WARN level, if we're changing it to not even print the characters?).

Yeah it makes a lot of sense to log OSC sequences, including unknown ones. But probably not one log statement per character, since the sequence only makes sense as a whole (and it can be hundreds of characters long). So I'd log it only after receiving the sequence terminator.

a new configuration parameter for the protocols that use the terminal that allows toggling whether or not the characters are ignored

I don't expect that it will be a good idea to a add this config option, because turning it on will randomly break applications, and it's not even really useful for debugging, because the prefix of the sequence is consumed. So if you run a program that sneakily prints some escape sequence that's not supported by the terminal; printing a suffix is

1. noisy (because in these cases, users don't actually care that the escape sequence is not supported because it was printed behind their back)
2. not even enough to identify the sequence, because the prefix is missing, so it could be OSC, DCS, APC etc. I don't think any terminal has this type of option.

[krobelus]

let me know if this patch works for you, I think it's an obvious fix. I guess we can add logging too but I'm not sure if that should block this fix