On 7/19/2018 8:55 AM, Joseph Canedo wrote:
Thanks.
The problem happens in a case where there are no paragraph split between 2 pages. All pars are entirely either in page 1 or page 2 (these numbers are not actual page numbers, but it’s for clarity).
I am trying to create a MWE but it’s not trivial. The situation as far as I understand it is like the following :
I’ve added tracers for vspacing.synchronizepage function.
When page 1 ships, there are 2 paragraphs (1 + 3 lines) still pending because they do not fit in page 1 so they will be shipped in page 2 (I think that what is in texlists.page_head variable). Hence tex.prevgraf which was 0 is set to 4 (see this from traces in output log).
what do you mean with pending ... tex doesn't look ahead, only back you probably need to look at the delta of pagegoal - pagetotal and evide that by the line height .. probably good enough for text without whitespace even then, hyphenation penalties can make a page one or more lines shorter so you also need to disable some of that
vspacing > page 4, prevdepth 2.56577pt => 2.56577pt, prevgraf 0 => 4
vspacing > list glue:topskip hlist:line penalty:userpenalty penalty:userpenalty glue:baselineskip hlist:line penalty:linebreakpenalty glue:baselineskip hlist:line penalty:linebreakpenalty glue:baselineskip hlist:line glue:userskip
Next paragraph after the 2 pending pars I refer to above has a non zero \hangindent and \hangafter (set to -1) and when it starts being processed (in horizontal mode) prevgraf is already 4 so I think this cancels out that hanging (that’s the effect I see in output, there is no indent in 1st line of 3rd par), as IIRC TEX considers par has already 4 lines but in actuality it has 0. If I read correctly the TeXBook, prevgraf value has some impact on \hangafter / \hangindent
you never really know what you look at: tex collects stuff and triggers the output routine when it thinks it makes sense in context there is also the rather complex vertical spacing mechanism that needs to look back and forward while looking back can interfere with page breaks (and/or retrying) these mechanisms evolved over time and are still nto perfect
The value of \prevgraf affects line breaking only when TEX is dealing with nonstandard \parshape or \hangindent.
no, it's independent ... just th enumber of lines (hboxes) that make a paragraph after it's broken into lines hangindent/after/parshape don't look at real dimensions, they only count lines this is why a side float mechanism is so complex ... lots of border cases and trickery around page breaks (because not only the size of what goes side is important but also the preceding and following space, the height and depth of lines etc
Joseph
*De : *Hans Hagen mailto:j.hagen@xs4all.nl *Envoyé le :*jeudi 19 juillet 2018 00:00 *À : *Joseph Canedo mailto:josephcanedo@gmail.com; mailing list for ConTeXt users mailto:ntg-context@ntg.nl *Objet :*Re: [NTG-context] Issues with vertical spacing in simple case
On 7/18/2018 10:33 PM, Joseph Canedo wrote:
Thanks Hans.
Picked up today beta and some code I have relying on value of
tex.prevgraf starts breaking. This is lettrine related code that sets
hangafter for 1st paragraph and eventually subsequent if lettrine is
larger than 1st par.
I see the function vspacing.synchronizepage() sets prevgraf value I
guess to avoid bleeding also paragraph lines.
Is there any possibility to get original unchanged tex.prevgraf value,
possibly using some callback or some other tex.<Something> ?
you cannot rely on prevgraf crossing pages (so you get e.g. 8 at the top
of a page with 5 lines
where/how does it fail
Hans
-----------------------------------------------------------------
Hans Hagen | PRAGMA ADE
Ridderstraat 27 | 8061 GH Hasselt | The Netherlands
tel: 038 477 53 69 | www.pragma-ade.nl | www.pragma-pod.nl
-----------------------------------------------------------------
-- ----------------------------------------------------------------- Hans Hagen | PRAGMA ADE Ridderstraat 27 | 8061 GH Hasselt | The Netherlands tel: 038 477 53 69 | www.pragma-ade.nl | www.pragma-pod.nl -----------------------------------------------------------------