Re: new version of Lout imminent

[kahl]

Dear Jeff,

 > I am getting together a new version of Lout.

This is good news! Thank you!

 >    Fixed bug with large page sizes.  Lout stores object sizes in 32-bit
 >      integers, but smaller numbers (gap widths, shifts, character metrics
 >      etc.) in 16-bit integers.

Since gaps can be almost as wide as @Wide, shouldn't gap widths perhaps
have 32 bits, too?


If you send me your current state, I can try it on an interesting case:

What is -ddd supposed to do? (I did switch it on.)
Currently I am consistently getting a segfault on the second run,
and I have never seen any tracing output from -ddd.
(I haven't even been able to figure out where in the code that command-line
 option is recognised.)

Currently I have a file where Lout crashes immediately,
i.e., before any visible PostScript output (after rm *.l[di]*),
| lout file "slides.lt":
|    337,1: unresolved cross reference @Runner&&following
| lout file "SlidesSyntax1.lt" (from "slides" line 381):
|      3,1: unresolved cross reference @LectureList&&syntax1
| lout file "slides.lt":
|    337,1: unresolved cross reference @Runner&&following
|    337,1: unresolved cross reference @Runner&&following
| 18736 Segmentation fault

(long before the first Haskell fragment)

but works with MALLOC_CHECK_=2 !!


And, worse: address@hidden {X}'' or address@hidden {X}'' makes lout hang
(even on the first slide),
while replacing this with address@hidden {N}'' or address@hidden {N}'' works.
(Took me ages to figure that one out...)

Even if this is a strange side-effect from one of my own definitions,
that's what I would expect tracing and debugging options like -ddd
to help me find out...

(Those segfaults and hangs can get in the way quite badly...)



By the way, I find it dangerous that some .ld files are valuable sources,
and some are intended to be invisible byproducts.


I have included the diff from the current state of my tree:

-           NoTitles.??       @Yield ??
+           NoTitles.??       @Yield "_"

This is essential for slides: The ?? is higher than the eventual
running header, and lets the otherwise fitting slide contents overflow.
In the worst case, I need hundred runs to converge...
The "_" solves the problem.

Most of my @Haskell problems are related to the fact that I
have not had enough time to try to understand prg2lout sufficiently
to fix them properly...

But there were some real problems in the last version:
* Comments and characters must not use the same function @PC
* ``@'' is an operator character, and is important for as-patterns:

  sortPair p@(x,y) =if x > y then (y,x) else p

* ``$'' is an operator character and needs to print as ``$''.
  Here is its prelude definition:

  f $ x =f x


makefile:  I do this in each version --- maybe a matter of taste:

! prefix=/usr/local/packages/lout-3.29wk
! BINDIR        =$(prefix)/bin
! LIBDIR        =$(prefix)/lib
! DOCDIR        =$(prefix)/doc
! MANDIR        =$(prefix)/man/man1



Hope at least some of this is helpful!


Wolfram

lout-3.29_wk_2004-10-21.patch
Description: Binary data

cprintWK.lt
Description: Binary data