On 25 September 2025 at 21:18, Christian Kastner wrote: | Control: severity -1 wishlist | | Hi Dirk, | | (sorry all for the duplicate mail, I forgot to CC the relevant addresses) | | On 2025-09-25 17:58, Dirk Eddelbuettel wrote: | > Thank you so so much for packaging llama.cpp and ggml. I look forward to | > working a bit more with it and seeing about possible R-based | > frontends. Discussing this with someone else I mentioned pkgconf / pkg-config | > and tried to check only to see that a) yes of course it ships a .pc file in | > the -dev package but b) it appears to be in a non-standard location: | | > It would be lovely if you could move the .pc files. | | keeping these out of the standard search paths was intentional, for the | same reasons that we don't ship the libraries in standard locations: | these libraries are not yet stable.
Hm. I do not think that is very clever. It basically just renders the library much less deployable. And I am not being argumentative here but why did you bother packaging it (which from the complexity of the upstream setup is surely non-trivial work) when at the end of the day you do not want user of the library use it? I am being serious here. I was planning to work "on top" and now I can't because I would have to re-invent library discovery on every possible distro or deployment. Strikes me as suboptimal, so vendoring remains the default. And I do not like vendoring. Hence the question of "why bother packaging" ? | Shipping them in private directories is our way of saying "you can use | this, but don't be surprised if things break between upgrades". | | The Policy only mandates this hygiene for the libraries themselves, but | it felt wrong to only hide them, only to expose them again through | standard development tools. | | Upstream is working hard on stabilizing though, just last week ggml | started experimenting with semantic versioning. I intend to help with that. Oh well. At least now I know. Feel free to close it then. Thanks again for packaging, even if it is way less useful to me than I anticipated. And thanks for the prompt reply. | PS: Note even llama.cpp itself uses ggml by means of these hidden | directories. I guess I could hardcode those paths if they were stable across deployments. In reality the library likely moves too fast anyway. Cheers, Dirk -- dirk.eddelbuettel.com | @eddelbuettel | [email protected]
