Created attachment 97535 [details]
autocorrect database

this is a followup of Bug 77508	where this weird autocorrect regression bug was initially reported and confirmed by other Windows users

STEPS TO REPRODUCE

- download the attached acor_und.dat file
- place it in LibO user profile under the "autocorr" subfolder
- open a Writer new document
- type 8 followed by space or any other typing character (i.e. 8, 8a, 81, 8!)
- you will see weird automatic corrections (i.e 8°, 8°a, 8°1, 8°!)

RATIONALE

basically in LibO 4.2.x the asterisk key is now interpreted as a wilcard in autocorrect replacement table unlike 4.1.x where it was interpreted as a normal typing character

this results in unwanted weird autocorrections in 4.2.x

so, if you had a 4.1.x autocorrect entry like:

8* --> 8°  (which is contained in the acor_und.dat file and can be shown in autocorrect replacement table in "Tools/AutoCorrect options/Replace")

the autocorrect will correctly change 8* to 8°

if you have the same 8* --> 8° entry in 4.2.x now it will change any typing combination involving 8 and any other following typing characters.

i.e.
8 [space] --> 8°
8a --> 8°a
81 --> 8°1
8) --> 8°)
8° --> 8°°

and so on....


so, any autocorrect entry inherited from 4.1.x in which an asterisk key is present, will result in unwanted and impredictable autocorrections in 4.2.x.

IMHO the autocorrect engine should not use wildcards because it breaks compatibility with previous autocorrect database and because wildcards may result, as in this case, in unwanted and impredictable autocorrections.