[NTG-context] Callbacks in LuaMetaTeX

Hi list, I've been playing around with some of the Lua callbacks in LuaMetaTeX, and I have a few questions/comments. Context: I'm writing a Plain/LaTeX/ConTeXt module called "lua-widow-control" that uses Lua callbacks to automatically remove widows and orphans from documents. The relevant Lua code is at: https://github.com/gucci-on-fleek/lua-widow-control/blob/master/source/lua-w... in case you're curious about how I'm using these callbacks, but you shouldn't need to look there since this email is self-contained. 1. In LMTX, calling "tex.linebreak" produces a luatex warning > tex: left parfill skip is gone warning. I don't get this warning in Plain LuaTeX, LuaLaTeX, or MkIV, so I think that it's specific to LuaMetaTeX. The LuaMetaTeX manual hardly mentions "left parfill skip"/"parfillleftskip", so I'm not too sure what to do to avoid this warning. You can reproduce the warning with this code (but any call to "tex.linebreak" also works): \startluacode function test(head) tex.linebreak(node.copylist(head)) return head end nodes.tasks.appendaction("processors", "after", "test") \stopluacode \starttext Hello! \stoptext 2. The LuaMetaTeX manual says that "pre_linebreak_filter" is called after the parfillskip glue has been added, but this doesn't seem to be the case. With LuaLaTeX/Plain LuaTeX, this is true, but the node list passed to the callback in MkXL is missing the parfillskip. (I don't have an MkIV installation to test here.) This MkXL code: \startluacode function test(head) local last = node.slide(head) print "START" for i=1,3 do print(last) print( node.type(last.id), node.subtypes(last.id)[last.subtype], last.width, last.stretchorder, last.stretch ) last = last.prev end print "STOP" return head end nodes.tasks.appendaction("processors", "after", "test") \stopluacode \starttext Hello! \stoptext produces: START nil : glue spaceskip> glue spaceskip 341648 0 384354 2258 : glyph unset> glyph nil 213792 nil nil 2521 : glyph unset> glyph nil 385140 nil nil STOP START nil : glue userskip> glue userskip 0 2 65536 2357 : glyph unset> glyph nil 385140 nil nil 2611 : rule strut> rule strut 0 nil nil STOP while this (presumably) equivalent Plain LuaTeX code: \directlua{ function test(head) local last = node.slide(head) print "START" for i=1,3 do print(last) print( node.type(last.id), node.subtypes(last.id)[last.subtype], last.width, last.stretch_order, last.stretch ) last = last.prev end print "STOP" return head end callback.register("pre_linebreak_filter", test) } Hello! \bye produces: START nil : glue 15> glue parfillskip 0 2 65536 298 : penalty 2> penalty linepenalty nil nil nil 82 : glyph 0> glyph unset 182045 nil nil STOP I'm not sure if this is a doc bug, an engine bug, or if I'm just doing something wrong. 3. The LuaTeX/LuaMetaTeX engine manuals say that for node callbacks, you can return "true" if you don't need to replace the passed node list. This works fine with raw LuaTeX callbacks and the luatexbase wrappers, but it doesn't work with the ConTeXt "nodes.tasks" wrapper functions. This Plain example works: \directlua{ callback.register("pre_linebreak_filter", function () return true end) } Hello! \bye But this MkXL example doesn't: \startluacode function test(head) return true end nodes.tasks.appendaction("processors", "after", "test") \stopluacode \starttext Hello! \stoptext I'm not sure if this behaviour is expected or not. Saving the node passed to the callback then returning it at the end works just fine, but being able to just return "true" would be slightly more convenient sometimes. 4. Would it be possible to add a ConTeXt-level interface ("nodes.tasks") to the "pre_output_filter" callback? Right now I'm using the low-level engine interface ("callback.register"), but this is pretty fragile since any format update could block this. Plus, it would be more convenient for me if I could register all of the callbacks the same way. The way I'm doing it right now seems to work well enough though, so this doesn't matter too much. 5. Off-topic, but is there a good way to trace node memory leaks? Documents larger than ~1600 pages run out of node memory when I load lua-widow-control, so I know that I'm leaking nodes somewhere, but it's pretty tricky to trace. Are there any "tricks" to find leaked nodes, or do I just need to watch my allocations very carefully? Sorry if any of these points don't make much sense or have already been answered elsewhere. Thanks in advance. -- Max Versions used for test documents: ConTeXt ver: 2022.04.20 19:18 LMTX fmt: 2022.5.1 int: english/english This is LuaTeX, Version 1.13.2 (TeX Live 2021/W32TeX)