Hi Rene,
Am Tue, Jan 10, 2023 at 08:30:52AM +0100 schrieb Rene Engelhard:
> >> The problem per se is that you copy private stuff over.
> >
> >I'm aware that this is not a nice solution - I considered a bit better
> >then keeping a full code copy of hunspell which is shipped by upstream.
>
> Oh my... Yeah, it probably is. If you filed a bug with a wish for a
> libhunspell-private-dev or somesuch (as e.g. poppler or Qt do) I am not sure
> I'd have done that but that'd probably be the cleanest solution. (Though
> given the waiting time in NEW for simple binary package additions I'd now not
> like to have to do it..
> )
I have no experience how frequently those private headers are changing.
I'd say: If we run into similar trouble again we could consider a
cleaner solution - but for the moment it workd somehow.
> >On the other hand this would saved us some work now.
>
> Maybe, yes, but...
:-)
> >After merging your MR the package builds. However, I get two failures
> >when running its test suite:
> >
> >
> >══ Failed tests
> >
> >── Failure ('test-encodings.R:16'): Dictionaries are found
> >─
> >hunspell_info("en_US")$wordchars not equal to "’".
> >1/1 mismatches
> >x[1]: "0123456789’"
> >y[1]: "’"
>
> Groan. Looks like it needs some specific stuff in the dictionaries it has
> which are now used from the system. I don't think it should not do that,
> can't the tests be adapted?
Possibly. I'm fine with skipping these two tests for the moment.
I'd assume that upstream will stumble upon this once the package
itself is upgraded to hunspell 1.7.2.
Kind regards
Andreas.
--
http://fam-tille.de
