CIDER locks up when printing large strings

[arrdem]

I have of late sacrificed many Emacsen to print lockup. Either due to nREPL network transport times or Emacs' known performance issues with long lines it's possible for a CIDER repl to freeze Emacs entirely such that it's not even possible to (kbd C-c C-c) or (kbd C-c C-b) to abort the print in progress. Killing Emacs is (obviously) a huge cramp on my workflow and it'd be awesome if this didn't happen.

<3

[bbatsov]

Might be related to #228. A reliable repro case and some quality time with the debugger/profiler should solve this.

[Malabarba]

@arrdem About the keys you mentioned, did you try C-g too?

@bbatsov Emacs is notorious for being slow at font-locking long lines. I'd say the first thing to try is to not font-lock the output string if it's very long (don't know if that's been tried yet).

[arrdem]

@Malabarba I would agree that this is more than likely Emacs' various long lines issues more than CIDER on further review. C-g locks up when I encounter this.

[bbatsov]

I don't think the problem's just font-locking related, although it's likely a good idea to disable font-locking for bigger results. There's something pretty wrong happening in the REPL with respect to output - one bigger result/output and it becomes nearly unusable. And this has been happening before the introduction of "fancy" font-locking for the results and the output.

[expez]

About the keys you mentioned, did you try C-g too?

I have previously managed to stop this by spamming C-g along with cider-interrupt, but it took a long time before it registered my desire for that process to just go off and die.

[bbatsov]

@arrdem Btw, you do experience this problem only in the REPL or for interactive evaluation results as well?

[arichiardi]

Confirm this

[bbatsov]

Just noticed there's some fresh profiling data here. Seems our custom REPL implementation sucks big time...

[zarkone]

I don't think it's about REPL, it's the same emacs long-line problem, or somehow connected with it. REPL buffer inserts the result with no new lines. For example, if i do cider-eval-last-sexp, i could spam C-g a few times to interrupt the process. But If I try cider-eval-last-sexp-and-replace, the situation would be the same, like i eval it in REPL buffer -- can't interrupt... I've tried to slurp 1,5M binary file, and after waiting some time with frozen Emacs i've seen the result -- long-lined result in REPL, and Emacs lagging because of it. So, the problem is reproducible not only in the REPL buffer.

I'll try to test interrupting with big text but without long lines.