Open u-fischer opened 4 weeks ago
I suppose we need to decide whether to support \endfirsthead
coming later (by which time the \endhead
rows have already been boxed). You could pick this up on a later run, or make that an error and specify that \endfirsthead
has to come first if tagging enabled,
Don't we need a label anyway? I mean when the first row of the \endhead
is seen the code can not know if that is a normal row which should be tagged or a \endhead
row that should be ignored. The alternative would be to stash all row structures until the end of chunk and then to decide if one use them or drop them but I would have to test that first.
hmm ok so fix up on second run... In the v5 experiments there is an option planned not to store the headers in boxes but (like supertabular) as macros, people often want marks or column sums etc, and forcing a fixed box has issues even without tagging.
well actually there is already code to handle this. It only didn't work because of https://github.com/latex3/tagging-project/issues/554
If a longtable uses both
\endfirsthead
and\endhead
additional (empty with luatex, with wrong content with pdftex) row structures are produce and the row count is wrong:We need a way to suppress the tagging of the \endhead and suppress its row count if there is \endfirsthead.