Hi all, Sorry, I did not get this email in my inbox, probably due to the way how BTS handles reassigns. So I found it only now while looking for other Sphinx bugs.
On Thu, Jun 25, 2026 at 06:34:01PM +0200, Niels Thykier wrote: > Hi, > > The `dh_sphinxdoc` has code that conditionally emits `sphinxdoc:Built-Using` > based on whether it is present in `d/control`. See > https://salsa.debian.org/python-team/packages/sphinx/-/blob/debian/master/debian/dh-sphinxdoc/dh_sphinxdoc#L552 > > This is certainly not an optimal interaction (between the compat 14 feature > and this conditional substvar). But I cannot make a judgement call on when > `sphinxdoc:Built-Using` should be emitted. Though, I suspect #1136263 is the > cause (it links to the commit that made the substvar conditional). Absolutely, #1136263 is the reason. I believe ${sphinx-doc:Built-Using} was a bad idea. It was meant to be used in rare occasions, but debhelper prints a warning when a substitution variable is defined but not used, so a lot of packages started to use it without a real need. And it created unnecessary coupling in the testing migration process: Britney does not allow a package built with a new Sphinx version migrate before Sphinx itself. And debhelper 14 was one of the reasons for this change: if I did not do it, debhelper would add this field automatically, even to the packages that did not use it before. I definitely want this field to be opt-in, and I don't want debhelper to warn when it's not used. Doko even suggested to me that I do MBF for all packages using it currently. > Perhaps, it can use `Static-Built-Using` instead and then packages that need > it for licensing can promote it to `Built-Using` (via `Built-Using: > ${sphinxdoc:Static-Built-Using}`)? This is an interesting idea. If that field does not affect testing migration, then it's a better choice than normal Built-Using. However, it is not documented in the Policy, and its name suggests that it's intended for statically linked binaries, which Sphinx has little to do with. And before I do that, I want to know whether we still need this field at all, and where does the requirement for it come from. I am CCing Sean Whitton, who requested this field originally in #836248. Dear Sean, do you know if the reasons for needing this field still stay? If yes, do they apply to all packages, or just packages with a specific license? And will Static-Built-Using field be enough? -- Dmitry Shachnev
signature.asc
Description: PGP signature
