Moving dictionaries to build from a single source package

at libreoffice-sig we started to think about moving several hunspell-*, hyphen-* and mythes-* language packages from building as separate source packages to a single source package built from libreoffice dictionaries.
Our thought is that way would be much more simpler to update and manage them.

I would like to gather some opinions on this change before submitting as a Self-Contained Proposal (I think I’m still in time for F39), especially from @pnemade who seems to be the main admin of many of the current language packages.

The plan is to have three main packages named hunspell-dictionaries, hyphen-dictionaries and mythes-dictionaries and move as much as language possible to build from there and retire the old singleton packages. A draft start is at Build 6132492 in mattia/Testing
Note that not all language packages currently available in Fedora are present in the LO sources, so those they will remain untouched. On the other hand, there are several other languages in LO sources which are not available in Fedora, so we could easily add them now or in future.

As an alternative, we could create a libreoffice-dictionaries and build all the hunspell-*, hyphen-* and mythes-* subpackages from there… what do you think?

1 Like

I will prefer to keep all those packages separately. Benefits more visibility of the dictionary and its URL and files getting installed. Also, upstream versioning of those dictionary sources.

Note that the proposal is to generate individual (sub-)packages, but from a single source package, which makes it much easier to do rebuilds and apply uniform packaging standards.

Alternatively, would you be willing to take up all those packages so that they are in one hand at least?

As for the versioning: Are they versioned individually upstream? Are they not distributed as part of a specific LO version? Reminds a bit of our TeXlive packaging: TL version vs individual component’s version.

I know this new proposal generate subpackages but I still can’t find benefit of this new approach. I am so used to with maintaining individual dictionary packages for many years that I don’t feel need to combine them.

I think I taken ownership of most of hunspell-* and hyphen-* packages.

I always preferred individual sources by recognized country specific institutions providing these dictionaries and considered them as upstream. Only where I did not find such upstream, I fall back to LibreOffice github dictionary source.

AFAIK, the libreoffice-dictionaries repo started from combining all upstream multiple sources into one, but many (all?) languages are now updated only within the LO dictionaries repo.
For example, I’ve recently updated the Italian language which upstream was dead, while LO was recently updated.

There are 22 hunspell and 3 hyphen packages that uses URL as or 33 hunspell and 3 hyphen packages where Source: is

Then there are 11 hunspell packages that uses Source: as

Marathi(mr_IN) and Malayalam (ml_IN) people prefer their own hosting.

I’ve taken over 29 hyphen packages when they were orphaned and nobody wanted them. If you maintain most already and prefer to continue with individual packages then this proposal is futile at least for hyphen*, but I’d ask you take over or co-maintain all of them then. Should I add you and/or the SIG, or simply orphan the packages? Our tooling is rough around the process of orphaning/comaintainer notices.

I was waiting patiently for some decisions to happen before I own most or all of them. Now that I can have some answer I can say, I can own all hunspell* and hyphen* packages in Fedora.
I request to keep libreoffice SIG as CC member to all these dictionary packages and you can also add me as co-maintainer to required/remaining hunspell-* or hyphen-* packages.

Just in time I remembered we can “give” directly rather than orphan. That saves you the time to pick them up. Done! (given to pnemade, commit lo-sig)

Thanks for taeking cayre of Fedorah speliingk!

1 Like