On Tue, Oct 27, 2026 at 12:41:15PM +0200, Eli Zaretskii wrote: > > I attach the .texi and .info in case you would like to test it. > > This is very hard to support properly in Emacs. The reason is that, > once the node name is gleaned from its reference, either in a menu or > in a cross-reference, the Emacs Info reader strips the ^?..^? quoting > from it. That's because the node name is then used in gazillion > places: for entering it into node history and comparing with other > node names, for showing it on the mode line, the header line, and the > breadcrumbs, for displaying it in a tooltip when the mouse hovers > above the reference -- and, of course, for going to that node when the > user types RET or clicks the mouse on the reference. (Keeping the > quoting instead would require to remove them each time the node name > is used as a string, without interpreting it, which is both much more > complex and would create a lot more temporary strings all over.) And > when the quoting is stripped, there's nothing to tell the rest of the > code that the parenthesized part of the node name is not a reference > to a file name of an external manual, a convention that is deeply > entrenched in how an Info reader handles node names.
Indeed, if the quoting is stripped, this is difficult. One option, maybe, could be to add the current manual as a leading manual, I have seen that when testing some commands with the standalone Info reader, such that the reference becomes, for example (if in the f.info manual): (f)(toto)titi instead of (toto)titi > Moreover, I don't see how such node names can be supported > consistently in any Info reader, because at least when the user tells > the reader to go to a node by name, typing "(foo)bar" at the reader's > prompt _must_ be interpreted as node "bar" inside file "foo", or else > we will break compatibility with a long-standing tradition in Info > usage. I tested with the standalone Info reader, and indeed g and G do not work with a parenthesized name, or, rather, it is proposed for completion, but interpreted as an external node, as you say. However, it works well in for all the other commands I tested, such as m, f, M-TAB, i. > And it's unreasonable to ask the user to type the quoting > characters to prevent such interpretation. I tried to add DEL with C-q in the standalone Info reader, and it did not seem to work. Regardless, I agree that it is unreasonable to ask the user to type the quoting characters (but it could still have worked, can be useful for testing). I tried g(f)(toto)titi and it works, but there is no completion. > So I think we do not need to support such node names. If some manual > uses them, they are on their own. My feeling is that with the standalone Info reader, it is not so bad. I do not thing that G is often used, not having the possibility to go to some node with g without prepending the manual name is clearly annoying. Yet, it is not a show stopper in my opinion. Let see what Gavin says. -- Pat
