On 10 December 2022 at 22:38, Nilesh Patra wrote: | On Sat, Dec 10, 2022 at 07:56:52AM -0600, Dirk Eddelbuettel wrote: | > | > On 10 December 2022 at 09:07, Peter Green wrote: | > | Source: tiledb-py | > | Version: 0.18.2-1 | > | Severity: serious | > | Justification: rc policy - "Packages must be buildable within the same release" | > | x-debbugs-cc: e...@debian.org | > | User: debian...@lists.debian.org | > | Usertags: edos-uninstallable | > | | > | tiledb-py build-depends on libtiledb-doc, which is no longer built by tiledb since | > | version 2.13.0-1, this removal is no mentioned in the changelog, so it's not clear to me | > | if it was deliberate or not. It is still present in unsable as a cruft package, but is | > | completely gone from testing. This means that tiledb-py in testing is in violation of | > | the rc policy. | > | > Good catch but that was in fact deliberate. | > | > The build (of a package I inherited / adopted) was giving me fits, and I as | > maintainer have decided to follow a) upstreams preference for documentation | > on the websites and b) simplify the build. | | While this is an acceptable stance, I'd really prefer if you consider to keep | vendoring the documentation. I have seen a number of bug reports and also heard | from many people that they'd like to have a copy of documentation offline as well, | as it a) enables to work when internet is spotty for them b) Look up everything | offline instead of the online source as the docs contain API that correspond to | the relevant version. | | I understand that vendoring documentation could be extra work, but if vendoring it | is not a source of nuisance for each and every update, and the build rules are constant | then I don't see a lot of issue with it. For your case, did you have any particular | issues/build failures while vendoring the documentation? | Also, I know that you understand tiledb far better than I do, but still I'd like to offer my help | for this issue, should you like it, and if you help me understand where exactly it crossed | the threshold for maintainer-time-well-spent. | | > We should adjust tiledb-py | | This is easy enough to do, which would mean removing doc package from tiledb-py as well. | But again, I'd like to do this only after I hear back from you.
It's complicated. You and I may both have too many packages to dive into why this one suddenly fails on docs. To me, the library part matters more. We could always build a -doc package off the same sources, or debug the build, or ... but I do _not_ think we should hold the library package back for this. So I kindly ask you to adjust tiledb-py to unlock the tiledb (source) package for no longer shipping libtiledb-doc. | > (which needs an update for the now released 0.19.0 | > anyway, and had skipped minor release 0.18.3 which is ok) accordingly. | | Thanks for the ping. I work on hunderds of packages and I tend to skip updates so this | is helpful. It's a titch under two hundred for me and I keep an eye on my QA page every day or other day which works well enough. Plus, many of them are r-cran-* and I see via CRANberries [1] when those change upstream and quickly update those. Dirk [1] As RSS or tweets or here in html: https://dirk.eddelbuettel.com/cranberries/ -- dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org
