Hi all,


At the last Stylesheets meeting, we checked how the command "xmllint --format -" that is run when testing the HTML transformations adds the XML declaration to the output file (which was a bit counterproductive because the XSLT code was explicitly demanding the omission of the declaration).

Now I’ve run into another problem that is also generated due to the xmllint command: it outputs XML entities when a literal string is expected. I run the script without xmllint and the output is the expected one so it seems that it is another of the modifications done by this program. This issue primarily affects the quotation marks that are added in the HTML output using CSS. See diff:


17c16
<     content:"&#x2018;";
---
>     content:"‘";
20c19
<     content:"&#x2019;";
---
>     content:"’";
23c22
<     content:"&#x201C;";
---
>     content:"“";
26c25
<     content:"&#x201C;";


My guess is that this wasn’t an issue before because xmllint had as an input a file with the XML declaration. Now it finds a file whose first line is the HTML doctype declaration and thus adds these modifications.


Considering that in the HTML transformation Saxon already has the command to indent the results, my proposal is to modify Test/Makefile and delete the "xmllint" command ONLY for the transformations to HTML. Would this be a viable solution?



Best,


H.



Helena Bermúdez Sabel
Chercheuse FNS senior
Institut des sciences du langage
Université de Neuchâtel