Re: Huge auto-fill-mode bug intro'd in 20.0 still not fixed in 21.2.19

Hrvoje> "Dan Harkless" <dan-xemacs(a)dilvish.speed.net&gt; writes: Hrvoje> At least part of the problem is that there is (or should Hrvoje> be) no such thing as a "Japanese buffer". Any buffer can Hrvoje> contain any character. XEmacs does not yet know anything Hrvoje> about specific languages, either. No, but obviously modes that do text-filling should know about them. Dan> We'll revert to the state of affairs when things worked for Dan> European languages, and then we'll accept correct fixes for Dan> Japanese. OK. Since Dan has provided an algorithm to replicate the problem, it should get fixed soon enough. Don't forget put "there are changes in this version of XEmacs that cause potential data loss for Japanese (at least) using kinsoku processing" in the ChangeLog entry, NEWS, and PROBLEMS. Dan> I don't have an xemacs built with mule, so I can't try it, Dan> but do you get lines of (fill-column + 1) columns if you type Dan> a space at fill-column in Japanese too? OK, I see this. So it's not the `(if (featurep 'mule) (kinsoku-nantoka))' calls. Pity. Dan> If so, it seems like the current code is wrong for all Dan> languages -- what about my previous suggestion that the right Dan> thing to do if you need to preserve spaces is to have Dan> auto-fill actually go back and break at the _previous_ legal Dan> fill boundary? Then the unbreakable text at the end of the Dan> line and the space after it would go down to the beginning of Dan> the next line. This would not be acceptable in Japanese. I think it would be pretty unacceptable in English too (people will go nuts with three-letter words that don't fit into four letter spaces at the end of the line). Dan> There could even be a variable like Dan> [auto-?]fill-preserve-spaces or something like that that you Dan> could set in any language that would cause that behavior. If Dan> it wasn't set, the old correct (Western language) behavior of Dan> changing the space to a newline would apply. Unfortunately, people who write in Japanese often need both behaviors in the same buffer (as Hrvoje points out). Steve Baur suggested that we could do a charset test on the character at a candidate point; I think that's good enough in general. -- University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN Institute of Policy and Planning Sciences Tel/fax: +81 (298) 53-5091 __________________________________________________________________________ __________________________________________________________________________ What are those two straight lines for? "Free software rules."