Re: Scan of regexps in emacs

[Mattias Engdegård]

11 mars 2019 kl. 04.37 skrev Paul Eggert <address@hidden>:
> 
> Clément Pit-Claudel wrote:
>> On 10/03/2019 22.45, Paul Eggert wrote:
>>> -            (while (string-match (concat "\\(^" comint-prompt-regexp
>>> +            (while (string-match (concat "\\(" comint-prompt-regexp
>>>                                           "\\)\\1+")
>>>                                   string)
>> I think your change altered the meaning of that regexp.
> 
> Yes and no. Yes, it altered the meaning of the regexp, but no it should fix a 
> bug rather than introduce one because comint-prompt-regexp in practice always 
> seems to be anchored to a line start. For example, comint-prompt-regexp 
> defaults to "^", which meant that the above code's entire regexp was this:
> 
> \(^^\)\1+

One way would be to replace the "\\(^" prefix with "\\(?:^\\)\\(", which should 
work whether or not comint-prompt-regexp begins with a ^. However, as you say, 
comint-prompt-regexp always seems to include the ^ and I think it is used 
elsewhere with the tacit assumption that it does, so the committed change 
should be fine.