Line disappears

[ralsina]

\begin[class=book]{document}
        \set[parameter=document.parskip,value=12pt]
        \set[parameter=document.parindent,value=0pt]
        \font[family=DejaVu Sans,size=10pt]{

\skip[height=0pt]\glue[width=0em]This is a second line block.\glue[width=0mm plus 100%fw]\break
% \glue[width=0em]\glue[width=0mm plus 100%fw]\break
\glue[width=0em]Blank lines are permitted internally, but they must begin with a "|".\glue[width=0mm plus 100%fw]\break

fooofoo

}\end{document}

If you uncomment the commented line, the line saying "Blank lines are permitted internally" disappears from the output PDF.

[alerque]

Manually ending every line with a stretchable glue is not a good way to achieve left justification. That's part of what's throwing you off here. You should either use the raggedright environment or set the document.rskip value manually. This achieves roughly the same thing as your manually placed glue but tells SILE a lot more about when to use it and when it can be discarded. In your case one of those glues actually needs to be discarded but it can't the way you've put them in.

Likewise if you want space at the start of a line you should probably be setting document.skip instead of stuffing glues in there. There are exceptions of course, but you'll get a lot fewer unexpected results if you don't try to do things manually that have other mechanisms with the edge cases worked out.

[alerque]

I messed around with this a little bit. The actual bug here can be boiled down to just this code:

\begin{document}
\raggedright{
Line 1\break
\break
Line 2
}
\end{document}

Using break on an otherwise blank like eats the next line.

That should probably be fixed. That being said, I suspect you're really looking for a different tool. The \break function should be used sparingly. You probably want to use paragraphs in most cases and setup your paragraph spacing to do what you want it to do. A blank like between two lines of text is usually indicative of a paragraph break, so use the tool that was meant for that job.

Additional notes on your code...it's best not to include extra whitespace such as indenting the \set commands at the top of your document. Putting those at the start of the line will help keep the output queue from getting cluttered up with things that it will probably discard. Also note the \font command can be used to set the current typesetter's font without having to wrap the content in the tag. This is probably preferable in the case of setting the document's default font.

[ralsina]

In his specific case it's a special sort of paragraph where line breaks and initial indents on each line are to be respected. It's what the definition asks for, so it's what has to be done to be compliant, so just setting it to raggedright would do totally the wrong thing, because it leaves control of line breaks to sile.

[ralsina]

Sadly, I am sort of convinced that sile is just too buggy for real use at this point, and tables seem to just not work, so I may have to freeze my project for a while.

[simoncozens]

Sounds like what you want is verbatim. But sure, with tables not working this might not be the right project for you at this time.

[ralsina]

Yes, a lot like verbatim, except that it's not monospaced and there are some minor corner cases.

[Omikhleia]

First, let's go for a simpler MWE:

\begin[papersize=a6]{document}
\neverindent
\nofolios
Line1\break
\break
Line2

Line1\break
\glue[width=1cm]\break
Line2

Line1\break
\kern[width=1cm]\break
Line2

\end{document}

Current output consists in "Line1" paragraphs only = every thing else including "Line2" disappeared.

The code responsible for that disappearance is here:

sile/typesetters/base.lua

Line 730 in 5bc372a

if seenHbox == 0 then break end

It this form, it can be traced back to ad4addc from Feb 21, 2014 (< SILE 0.9.0), or possibly even earlier, but at least that commit shows a removed comment that might have explained what was intended here:

XXX TeX would toss initial line glue. We don't, because we'll often put it back again. Instead we ignore it when we call the output routine. This may not be correct.

Whatever it was supposed to do, it doesn't lead to the expected result, so indeed this is a bug.

Removing that line leads to, in each case:

• A blank line <-- EDIT or that might have been the intent here, looking at ad4addc's log ("fix the blank line issue ..."). Dunno, but more likely than "glue tossing".
• The "Line2" line

[Omikhleia]

Additionally, due to the condition setting seenHbox, this is also the offending code that caused info boxes after an empty line to disappear in #1707 and it explains why adding an empty hbox in #1751 worked...