Re: Reverse Patch for Khmer?

Thanks that helps me understand what you meant.


Some additional thoughts:

I then assume that _some_ user of the existing version would be
> unhappy of loosing completely a functionality if that felt that worked
> 'well-enough' for them... and _they_ would file a regression bug if we
> rollback that change ?
There are two uses in LibreOffice for the ICU Break Iterator for Khmer. One
is line-breaks, which I am not saying needs to be removed - while not
perfect, it is usable, and is easy to correct when the line is broken in
the wrong place by adding a joiner character.

The only "feature" that would be removed is the use of the break iterator
for checking spelling. And since by using the break iterator for checking
spelling, it renders the spell checker quite inaccurate, no Khmer user
would be unhappy about this change because if they are using the spelling
checker they are going for 100% accuracy (no one wants to use a spelling
checker that only works half the time).
I should have more fully tested the patch before I said it was a good
addition (I affirmed it because I didn't fully understand the implications
- originally I thought there would be some way for the user to "see" where
the ICU break iterator was breaking the words, but as it is now, there is
no way for the user to see the breaks that are automatically inserted by
the break iterator - which is why I filed this bug:
https://bugs.freedesktop.org/show_bug.cgi?id=59447)


> At a proper version boundary, one could argue that it is necessary...
> but it is very harded to argue for that kind of changes between 4.0.n
> and 4.0.(n+1), since these dot release are meant to be 'safe' for
> user, as in very unlikely to do random behavior changes on you...

That makes sense, but because this patch renders the Khmer spelling checker
ineffective, I think it is worthy of reversing in a dot release.

There is another way to go about fixing this issue for Khmer (and it would
be a help to Thai users as well) if someone is willing to write new code.
 That way the current code could be kept as is, but two new features would
need to be added.

The most important would be adding the ability for the user to turn off the
ICU break iterator for their current CTL language as filed here:
https://bugs.freedesktop.org/show_bug.cgi?id=59448
and a secondary new addition would be making the ICU break iterator
zero-width spaces visible to the user so they can correct mistakes as filed
here: https://bugs.freedesktop.org/show_bug.cgi?id=59447

But since making such large changes will most likely take a long time, that
is why I suggested reverting the patch so that the ICU break iterator is
not used for spell checking with Khmer in LibreOffice.

Thanks,
Nathan