Watched word regular expression crash (2025)

Contribute Bug
,

AI-generated summary

A bug report was submitted regarding a crash caused by an invalid regular expression in the watched words feature. The issue occurs when enabling regular expressions for watched words that were initially set as wildcard expressions. Specifically, the regex *abc caused a crash because the * operator had no preceding token. RGJ provided a repro case and error details. pfaffman suggested validating regex before saving and adding exception handling. RGJ clarified the cause of the issue. sam proposed validating regex on save and marked it as pr-welcome. RGJ emphasized the need for exception handling around the regexp call. zogstrip confirmed that the issue will be fixed by a pull request.

1

Similar to this one

Repro

  • watched words regular expressions enabled
  • watched words tag action
  • regular expression *abc

Result

It is impossible to create topics or posts

RegexpError (target of repeat operator is not specified: /(*abc)/i)
app/services/word_watcher.rb:239:in `initialize'
app/services/word_watcher.rb:239:in `new'
app/services/word_watcher.rb:239:in `word_matches?'
lib/topic_creator.rb:214:in `block in setup_tags'
lib/topic_creator.rb:213:in `each'
lib/topic_creator.rb:213:in `setup_tags'
lib/topic_creator.rb:48:in `create'
lib/post_creator.rb:493:in `create_topic'

Cause

  • * means “repeat the previous token zero or more times”
  • there is no previous token
2 Likes
2

So really the regex should be validated before allowing it to be added? And there should be a rescue so an invalid regex can’t crash the whole world?

1 Like
3

What obviously happened here is that the watched words were regular watched words and * was the wildcard, and then watched words regular expressions enabled was enabled.

So this

1 Like
4

Well, now it’s obvious when you say it like that. :rofl:

5

Valildating regex in on save seems like a very sane thing to do.

putting a pr-welcome on this, but team will triage and decide if it makes sense to do this in the upcoming weeks to save us future support.

1 Like
6

That won’t prevent people entering “normal” wildcard expressions, and then turning on watched words regular expressions enabled, which is what happened here.

I think it just needs an exception handler around the regexp call

  def word_matches?(word, case_sensitive: false)
    options = case_sensitive ? nil : Regexp::IGNORECASE
    Regexp.new(WordWatcher.word_to_regexp(word), options).match?(@raw)
  end
2 Likes
7

Thanks for the report @RGJ, this will be fixed by

https://github.com/discourse/discourse/pull/35217

3 Likes
9

This topic was automatically closed after 10 hours. New replies are no longer allowed.