Hi,
As far as opcode regressions go I think the most obvious one is the group of
firstletter* and lastletter* opcodes. According to the documentation they used
to work on letters AND symbols. With the UEB patches applied they work on fewer
things, depending on what other signs you have defined and possibly on where in
a phrase these letters are encountered. At first glance the *word and *wordstop
opcodes appear to be more compliant with the documentation, but these differ in
other implementation details.
I think we should discuss these two groups of opcodes. My idea is that we
should deprecate one and more clearly define/document the other.
Another 'regression' with decpoint has already been addressed. To find other
problems I think it would be good to enlist the help of table developers where
possible.
Davy
-----Oorspronkelijk bericht-----
Van: liblouis-liblouisxml-bounce@xxxxxxxxxxxxx
[mailto:liblouis-liblouisxml-bounce@xxxxxxxxxxxxx] Namens Bert Frees
Verzonden: woensdag 5 augustus 2015 14:09
Aan: liblouis-liblouisxml@xxxxxxxxxxxxx
Onderwerp: [liblouis-liblouisxml] Re: Typeform weird bug
Hi Davy,
I already told you but I'm repeating it here for the others:
The main reason I've split off the table changes into a separate branch is to
discover these kind of regressions (if they are in fact regressions).
It's important that we try to make all tables including the old UEB table work.
First we mark all the possible regressions. Then we decide which ones are
actual regression and which opcodes are to be deprecated or redefined. Then, we
fix the regressions in the code, and we fix the tables that use deprecated or
redefined opcodes. Finally we apply the full UEB table changes from Michael.
Davy Kager writes:
I'm also seeing regressions in the UEB tables, but I'm ignoring thoseFor a description of the software, to download it and links to project pages go
since Michael has updated these tables so testing the old ones is a bit
redundant.
Davy
