Hi! I can't really squeeze out enough time or focus to really read through this thread fully, but I think my opinion as a Stupid User may influence this in the right direction here.
On Tue, Oct 04, 2022 at 11:55:41AM +0100, Colin Watson wrote: > On Mon, Oct 03, 2022 at 10:36:32PM -0500, G. Branden Robinson wrote: > > The problem is therefore that man(1) is running the preprocessors in the > > wrong order. This can be solved by pipelining eqn(1) before tbl(1), or > > passing the "-e" and "-t" flags to groff and letting it sort them out. > > This is because the preprocessor line at the top of the page is: > > '\" et > > man(1) deals with these in the order given. I would /never/ have guessed that. Bullseye man(1) (groff-base 1.22.4-6) says: -- >8 -- To contain a valid preprocessor string, the first line must resemble '\" <string> where string can be any combination of letters described by option -p below. If none of the above methods provide any filter information, a default set is used. -- >8 -- Nothing suggests this would have any bearing on the order ‒ these appear to function just like flags. > Should I have man(1) rearrange things into a canonical order? If so, I > guess I should rearrange everything rather than just tbl(1) and eqn(1), > to match what groff(1) does? tbl(1) says: -- >8 -- INTERACTION WITH EQN tbl(1) should always be called before eqn(1) (groff(1) automatically takes care of the correct order of preprocessors). -- >8 -- All signs (to me, noted dumbass; grain-of-salt) point to "just put the groff preprocessor-enabling flags you need here". It's not an outlandish assumption that they'd function the same? Or that if they didn't, and running the preprocessors out of order is known to break stuff, there would be a note of this? So idk, I did assume they functioned the same because nothing I saw says they wouldn't? So I think that'd be nice, yeah. As a point of reference, DCS has 454 pages (all of which manpages-l10n) of '\" et: https://codesearch.debian.net/search?q=%27%5C%22+et&literal=1 and the same of ^'\\".*e.*t (with a lot of normal text-comment false-positives): https://codesearch.debian.net/search?q=%5E%27%5C%5C%22.*e.*t&literal=0 > -- > Colin Watson (he/him) [cjwat...@debian.org] Best, наб
signature.asc
Description: PGP signature