Jeff King wrote: > On Sun, May 09, 2021 at 10:20:37AM +0200, Martin Ågren wrote: > > In general, I agree. I do think it's important that "most people > > contributing to Git", whatever that means, can build the documentation > > to check the part they're adding/modifying and not find it broken left > > and right. They would then (quite rightly) not even bother building it. > > Agreed. But I think that is mostly the case (asciidoctor seems no harder > to acquire on most modern systems than asciidoc; there are system > packages in most cases, and decent binary-package systems for both ruby > and python if you really need it). > > It does create a situation where people like Randall on NonStop might > need to do part of their dev work on another, more mainstream platform. > But I suspect that is already the case. Or use distributed tarballs with already built documentation. I don't see any big issue with cross-compilation though. > > When we looked at xmlto-less rendering around two years ago [1], we > > found various asciidoctor bugs up to and around version 2.0. We would > > likely need to require some >=2.0.x. The exact requirements will > > probably only become clear when someone really does the work. > > That does make things a little less convenient; Debian stable, for > instance, still has 1.5.8. And it has git 2.20.1, released at the end of 2018. I've never understood developers worried about how the bleeding edge would build in ancient platforms, when ancient platforms don't care about the bleeding edge. > It's not too hard to install an updated gem, but not quite as nice as > using the system package (it also makes things weird for building the > stable Debian package itself, which would want to rely only on other > packages; but of course any proposed change to the doc toolchain would > be for new versions, and would not get backported there anyway). Anyone trying to build git master on top of Debian stable 1. probably can live with the output of the current doc toolchain, and 2. probably doesn't exist. I'm not sure how much the git project gains by worrying about such a hypothetical person. > > I think what I'm arguing for is > > > > 1) switch the default to asciidoctor, > > 2) enable optionally using it without xmlto, > > 3) figure out what broke and fix it, and document which is the minimum > > asciidoctor version we're going to bother with for (2), > > 4) lather, rinse, repeat (3), > > 5) switch the default to not using xmlto, > > 6) drop the xmlto way of generating the manpages(?). > > I'm unclear when support for python asciidoc goes away here. Is it part > of step 6 (because it does not have another way of generating them)? Or > does it live on forever as a non-default legacy system? I'd prefer not, > but as long as we are clear about the primary target and leave it up to > people interested in the legacy to do the compat fixes, that might be > OK. How about we leave the legacy system in place as an alternative, and decide later what to do with it? -- Felipe Contreras
