Package compatibility caps



Why would this be capped? It was tagged after July 14.


like the strategy. i’t’s pragmatic, just like Julia



I just checked and it is, in fact, not capped so there’s nothing to be done here.


Sorry if I’ve missed a link in this thread, but is there a list somewhere of which package was affected? I looked for a METADATA PR and couldn’t find it.



Hi, what about the opposite case? I.e. I look at my packages and discover that an old version not meant to support Julia 1.0 (if fact has deprecation warnings for 0.7, so it would break in 1.0) is not capped. Must I report it or make some PR?


You can make a PR that adds upper bounds to the REQUIRE files that doesn’t support 0.7 / 1.0.


So what’s the best way to understand whether your package has such a cap on it?

Finding my way to:
indicates that this little package does.

And what’s the recommended way to fix this?
Push another minor version with the julia 0.6 2- REQUIRE trick of Stefan, and let Attobot sort it out? Or construct a hand-written pull-request to METADATA?


Does it work on 1.0?


Yep, tests AOK. I didn’t have to do much to get it working on 1.0 - just a few places with scope + waiting for Optim to become available.


Not much is still something. If you make a new release that claims to be 1.0 compatible that will be taken at face value.


As in, change my REQUIRE line to
julia 0.6 0.7 1.0
Or your julia 0.6 2- method?


As in just make a new release that actually works on 1.0 and leave the REQUIRE file alone. The julia 0.6 2- change is only required for versions of packages that were tagged before July 14th. You only have to change the require file in METADATA.jl if there’s an old version that works on 1.0 correctly and was erroneously capped. It sounds like that is not the case here since you had to make changes to get it working on 1.0. So just make a new release with julia 0.6 or julia 0.7 depending on what it actually supports.


Stefan, I did try to do this. I’ll try a semantic-versioning minor-number increment and see if that helps.


Something seems to be wrong with the process that generates the new registry from METADATA here, I’ll investigate. Sorry for the trouble and for not understanding what your trouble was before.


I was misremembering / misstating the heuristic before. I’ve added a rule that if someone registers a new version (since yesterday) that claims to support 0.7 or 1.0 then we just take that claim at face value.


Great! Many thanks Stefan. Happily installs + tests on 1.0.1 now.