On 2025-08-02 05:36, Blair Noctis wrote:
I have an ITP, #1101486, for the icu4x stack, which is basically the crates in your MR, then some, but as a single source package. It'd be a better organization, including the benefit of one source package building all binary ones, avoiding the turmoil of one-by-one uploads and builds, and avoiding 7 NEW trips. If you really want them to be in *now*, I won't stop you from uploading them. But since we are in the freeze, I guess it's not that urgent. So I'd appreciate some patience and, if possible, help with the single source package effort. You should be able to assume no practical difference between the 1:1 packages you already built, or the 1:N packages the drafted src:icu4x will build, in test builds of, in your case, matrix-synapse.
Blair, Thank you for packaging these---since they're all in the same upstream repository, I see no reason to have separate source packages (I'd also like to see this workflow represented in the documentation). I can certainly wait until we are out of the freeze, too. Does that packaging include zerotrie and also get zerovec-derive installable (which are also from the icu4x upstream package)? I also bumped tinystr (which apparently got absorbed into icu4x, too). How is that included/excluded in your uni-packaging? Especially given that it's already packaged. I also needed a litemap and twox-hash bump to have installable icu-segmenter-dev be installable. These other package changes are reflected in the MR (and were honestly most of the work). I can separate them out if I know what exactly from the upstream icu4x repo is included. Best, Antonio