On Wed, Sep 25, 2024 at 07:43:10PM -0400, Benjamin Kalish wrote: > Thanks. I can see it both ways, but still lean towards it being a bug. > Headings in HTML can contain HTML and I don't think the user has any reason > to expect that the content of a heading would end up anywhere else. If it > were up to me the deciding factor would be whether the literal content of > the headings need to show up in HTML attributes (and I don't see why they > would).
I had a look at the code, I could have missed some places. Attributes are based on Texinfo code at least in the following cases (the precise @-command being used possibly depending on split/unsplit or other customization): * @image <img> alt attribute * <abbr> title attribute based on @abbr/@acronym second argument * icon direction formatting <img> alt (based on @node or sectioning commands) * <meta> name="description" (based on @documentdescription or @*title and @node or sectioning commands). * <meta> name="keywords" (based on @*title and @node or sectioning commands). * <link> title attribute (based on @node or sectioning commands). * (button name or description as title or rel in href or link elements, but this can be ignored here as it is not based on user Texinfo code) I did not list the places where a file is expected (@image first argument, @verbatiminclude), but we can assume that users do not get the idea to use @inlineraw in those contexts. > If it is necessary, then yes, it is up to the user to avoid the use > of @inlineraw headings (and a warning would be most welcome!). @inlineraw with entities but no element is ok too, so a warning may be unnecessary. > On the other > hand, if it is not necessary and that content doesn't need to show up in > attributes, or if an escaped version could be used instead, I can't imagine any other change than removing the alt/title/metadata output (as described above). This is possible, as none of these attributes are strictly needed, yet the produced HTML is better with those, I think. Very motivated users could use the HTML formatting API to modify the formatting. Also note that this issue could happen for other formats, for instance in DocBook there could be similar issues with attributes. In LaTeX output, the commands that can end up in captions or table of contents cannot contain everything. > then the user > should be allowed to use @inlineraw here and a change to the code is > necessary to prevent noncompliant HTML output. It is difficult or even impossible to avoid noncompliant HTML output when raw HTML is output. There is this issue of expansion of raw HTML in attributes, but it is probably only one possible issue amng many others. How HTML generated from Texinfo code and raw HTML interact cannot be foreseen when parsing or outputting because HTML in raw HTML is not analysed (Texinfo in raw HTML is parsed as usual) but simply output unchanged. What about the following sentence in "Inline Conditionals: ‘@inline’, ‘@inlineifelse’, ‘@inlineraw’" node? Beware that in some command arguments, such as @node or sectioning commands raw @inlineraw text is output in diverse formatting contexts. Raw HTML could end up in HTML attributes, raw LaTeX could be used for table of contents. You should make sure that the raw code you provide is adequate in such contexts. -- Pat
