Page splitting: revamp, fix some issues

[poire-z]

(Upstream) lvstring: fix signed integer overflow
(Upstream) lvopc: fix memory leak
(Upstream) ldomDocument(): fix memory leak
(Upstream) lvtinydom: fix memory leaks
From buggins/coolreader#299

LVArray: fix indexOf() to work with any type

Fix oversight in our decb46c / buggins/coolreader@a112014#diff-f37b0ebc86a9c3391b92d47a8dce2d3cdb5cc270e2dcf07b4ef73af5e1d4af6f I guess,
(Pinging @virxkane for info).

LVDocView: fix m_twoVisiblePagesAsOnePageNumber uninitialised

CSS: line-break: accept "-cr-loose" to ignore no-break-spaces

Can be handy to counteract #423 (comment).

CSS: font-variant: extend effect of "normal" and "none"

See, and should allow closing koreader/koreader#7913.

lvrend: fix BlockFloat footnotes context line associations

Footnote links inside floats were previously associated to the line before the float, which could make them shown on the previous page if the float needs a page break to be shown without break.
We now delay these links associations, and associate them only when the float is fully passed by, so the footnote is never before the float.

In-page footnotes: support links inside inline-block

Was too lazy to handle it at the time, turned out it wasn't that complicated :)

Page splitting: revamp, fix some issues

Rewritten page splitter, which should be easier to understand and tweak.
The old code was quite tedious, trying to do everything in one single pass.
This new code first finds out group of consecutive lines that should be kept together (break avoid), and process each group when they end, and only then the footnotes they contain (with delay if they don't fit).
This fixes 2 small issues (see koreader/koreader#5031 (comment)):

• footnotes could slice a paragraph with floats on the line the footnotes happen; now, they are delayed till the end of the block, and the full block can go to the next page (with its footnotes) if it is taller than the current page.
• if a first footnote and its separator didn't fit on a page, a new page was made so it can be added, leaving some blank at bottom of previous page ; now, we delay footnotes in this case, allowing an additional line of the main text to be added if it fits.

---

This change is 

[poire-z]

No more CI checks :(
The links to previous CI checks have this on their page:
Since June 15th, 2021, the building on travis-ci.org is ceased. Please use travis-ci.com from now on.

[Frenzie]

Oh right, I forgot we were still using Travis here and there. I'll look into it.

[virxkane]

LVArray: fix indexOf() to work with any type

Fix oversight in our decb46c / buggins/coolreader@a112014#diff-f37b0ebc86a9c3391b92d47a8dce2d3cdb5cc270e2dcf07b4ef73af5e1d4af6f

Well, yes, it would be more correct.

[poire-z]

@Frenzie : do you want this PR yet around when dealing with #444 for testing ?

Also, 2021.07 is due by today - which I guess won't happen :) I'd like this PR to have at least a week of night(ly) life before it ends up in a release. So, should I bump it quick, or rather wait ?

[Frenzie]

Let's wait a week (unless you want to prep it, I can do some final editing and tagging pretty quickly most days). My cat is seriously sick. I'm in the waiting room atm with nothing to do but I didn't bring my laptop.

[poire-z]

Sorry to hear about your cat :/

Let's wait a week

You mean for the release, or for merging this PR ?
If for the release, then I guess we can merge this PR. Unless you need it if you plan to fix CI (which I guess you then won't soon - and that's ok!).
So, I guess I can merge and bump this?

[Frenzie]

I mean the release, yes.

[poire-z]

Mentionning a minor edgy-casy issue I'll be fixing, with screenshots for reference (easier done now while I'm at it, than later when I push a PR).

If a page ends with a link to a footnote that can be made in-page, and there is not enough room, the inpage-footnote is pushed to the next page. Nothing specific was done if the next page is in another "flow" (so, probably hidden): it would go into it.
With KOReader with "hide non-linear fragment", when on that main page (2nd page here) and taping to go to the next page, we would skip that hidden-flow page (and go to the 4th page here), and not see the inpage-footnote (and I was wondering: wtf, where is it ?! until I realized there is indeed a hidden flow page (the 3rd here), and I had to use Page Browser to jump into it and see the footnote):

The solution I guess is to get this:

with:

--- a/crengine/src/lvpagesplitter.cpp
+++ b/crengine/src/lvpagesplitter.cpp
@@ -932,6 +932,22 @@ public:
                 has_footnotes = true;
             }
         }
+        // Handle a little edge case
+        if ( cur_page_nb_lines == 0 && start_if_new_page >= 0 && (cur_page_nb_footnotes_lines>0 || !delayed_footnotes.empty()) ) {
+            // Empty page, but with footnotes or delayed footnotes not yet added.
+            // These footnotes are associated to the previous page's flow.
+            // If the lines we are about to add are from a different flow, we don't want
+            // these footnotes to be with them (or they could get hidden and skipped).
+            // If that's the case, add any delayed footnotes and create a new page.
+            LVRendLineInfo * line = lines[start_if_new_page];
+            if ( line->flow != prev_page_flow ) {
+                printf("line->flow != prev_page_flow %d != %d\n", line->flow, prev_page_flow);
+                flushCurrentPage();
+                // And call us again.
+                addLinesToPage(start, end);
+                return;
+            }
+        }
         #ifdef DEBUG_PAGESPLIT
             printf("PS: adding (%d+%d) to current page %d>%d\n", lines[start]->getStart() - cur_page_bottom,
                             lines_max_bottom - lines[start]->getStart(), cur_page_top, cur_page_bottom);

Feels a bit too easy :), so this will need testing, unless I was really prophetic when I wrote in the first post:

Rewritten page splitter, which should be easier to understand and tweak.