Instead of repeatedly reusing the file name 'PuzzleEngine.class' in
the main build directory, now each puzzle's NestedVM translation is
left in a separate subdirectory so that they don't collide with each
other. A bonus is that we don't have to rename the file back and forth
between the rule that builds it and the one that consumes it.
I've just upgraded my build machine to Ubuntu 18.04, which has come
with a version of javac that complains about both -source 1.3 and
-target 1.3. Both are surely pretty out of date anyway, so the path of
least resistance is to just increase them to the earliest version that
javac doesn't currently complain is deprecated.
The new client code is a lot smaller and nicer, because we can throw
away the col[] and num[] permutations entirely.
Of course, this also means that every puzzle that incorporated latin.c
now has to link against matching.c instead of maxflow.c - but since I
centralised that secondary dependency into Recipe a few commits ago,
it's trivial to switch them all over at once.
This is a dedicated algorithm for finding a maximal matching in a
bipartite graph.
Previously I've been solving that problem in this code base using
maxflow.c, modelling the graph matching problem as a restricted case
of the optimal network flow problem and then using a full-strength
algorithm for the latter. That's overkill, and also algorithmically
wasteful - the H-K algorithm is asymptotically better, because it can
find multiple augmenting paths in each pass (hence getting the maximum
benefit out of each expensive breadth-first search), as a consequence
of not having to worry about lower- or higher-value augmenting paths
in this restricted problem.
So I expect this algorithm to be faster, at least in principle or in
large cases, when it takes over the jobs currently being done by
maxflow. But that's not the only benefit. Another is that the data
representation is better suited to the problems actually being solved,
which should simplify all the call sites; a third is that I've
incorporated randomisation of the generated matching into the H-K
module itself, which will further save effort at each call site
because they will no longer have to worry about permuting the
algorithm's input - they just have to pass it a random_state and it
will take care of that internally.
This commit only introduces the new matching.c and builds a test
utility for it. I haven't yet migrated any clients of maxflow.
It's silly to have every puzzle using latin.c separately specify in
its .R file the list of additional modules that latin.c depends on, or
for that matter to have them all have to separately know how to adjust
that for the STANDALONE_SOLVER mode of latin.c.
So I've centralised a new pair of definitions into the core Recipe
file, called LATIN and LATIN_SOLVER, and now a client of latin.c need
only ask for that to get all the necessary dependencies too.
Also, while I'm here, I've moved the non-puzzle-specific 'latincheck'
test program out of unequal.R into the central Recipe.
Mostly just cribbed from the corresponding changes in PuTTY's
build setup, although since the two mkfile.pl scripts are not
_quite_ identical, I had to make a few tweaks.
It's a pain having it in a rule in Makefile.gtk, which isn't even the
recommended makefile these days - it can't be re-run conveniently, and
there's no way to parametrise it. Now it can be run no matter which
makefile you're using, and it lets you narrow down to a subset of
games (though not presets). Other options could easily be added.
A long time ago, it seemed like a good idea to arrange that binaries
of my puzzles would automatically cease to identify themselves as a
particular upstream version number if any changes were made to the
source code, so that if someone made a local tweak and distributed the
result then I wouldn't get blamed for the results. Since then I've
decided the whole idea is more trouble than it's worth, so I'm
retiring it completely.
[originally from svn r10264]
I'm going through all my projects and reworking them to avoid
depending on the monotonic integer-valued source control revision
identifier provided by Subversion, so I can migrate everything to git
without my builds and versioning breaking.
Puzzles's version number is now of the form YYYYMMDD.vvvvvv, where
vvvvvv is some string of source control information (currently still
the SVN-style "rNNNNN", but free to change in future). The date
provides monotonicity between my official automated builds, and the
second component is the one I'll be most interested in when people
send bug reports.
[originally from svn r10263]
Makefile.am, and there's a new mkauto.sh which builds a corresponding
configure script.
The old makefile has been renamed from 'Makefile' to 'Makefile.gtk',
indicating that the intended new _default_ approach is to use the
autoconf world. Makefile.gtk is provided as an emergency fallback in
case anything fails with the new stuff that used to work with it.
The new configure script does not support the same $(BINPREFIX) system
as the old Makefile did. However, as I understand it, it should be
possible to configure using --program-prefix="sgt-" (for example) and
then the binaries should all be renamed appropriately at install time.
The Makefile.am is quite painful. The Puzzles codebase relies heavily
on compiling individual object files multiple times with different the
cpp flags per build deliverable (program or library) and not per
source file. Solution: anything built with non-default compile options
has to go in its own little library. But that doesn't work either in
the general case, because as soon as you have more than one such
library linked into an application, Unix ld semantics bite you if the
objects in the libraries both refer to each other. So I ended up
building all those little libraries but not _using_ them - instead the
link commands for the programs needing those objects refer to the
objects directly, under the silly names that automake gives them.
(That's less fragile than it sounds, because it does _document_ the
names of the intermediate object files. But still, yuck.)
[originally from svn r9886]
the GTK front end, plus a 'make test' target in the GTK makefile which
uses them to automatically generate 100 puzzles for each game at each
preset configuration, test-run them back through the solver without
their aux_info to ensure that can cope, and produce an HTML box plot
of game generation times for each preset.
As part of this work I've removed the TESTSOLVE mechanism from r9549,
since the new --test-solve option does the same thing better (in that
when something goes wrong it prints the random seed that caused the
problem).
[originally from svn r9825]
[r9549 == 5a095b8a08fa9f087b93c86aea0fa027138b028d]
applets, here's an alternative webification in Javascript, using
Emscripten in asm.js mode (so that as browsers incorporate asm.js
optimisation, the game generation should run really fast).
[originally from svn r9781]
and Makefile.doc a command-line parameter 'BINPREFIX' which will be
prepended to all the game binary names. E.g. 'make BINPREFIX=sgt-' and
'make BINPREFIX=sgt- install', and correspondingly 'make -f
Makefile.doc BINPREFIX=sgt-'.
Also included in this commit by mistake, changes to singles.c to add
\n to the end of all its debug() statements. I meant to commit that
separately. Oops.
[originally from svn r9606]
front end using GNUstep, yielding a Unix program built from the same
code. Should make it easier to check OS X behaviour without having to
move as far as a Mac :-) However, it doesn't compile as is, so I'll
apply fixes to the code until it does.
[originally from svn r9498]
output (a comment from Recipe mistakenly got into the Unix
makefile); more importantly, lets mkfiles.pl run in up-to-date Perls
(where implicit split to @_ is now obsolete).
[originally from svn r8957]
(Missed off the explicit -o from the compile lines constructing
version.{i386,ppc}.o, causing both to be compiled as version.o and
dependent build steps to fail.)
[originally from svn r8933]
[r8931 == 36cee4e2796c23da15d3276e88416ad1ce035c4a]
of basically unchanged binaries due to the compulsory rebuild of
version.o. version.o now depends normally on version2.def, which is
constructed using much the same if statements that version.o used to
use, except that it's not overwritten at all if its contents don't
need to change.
[originally from svn r8931]
alongside the individual puzzle binaries, on Windows only. (MacOS
already has it, of course; Unix would require about as much work
again.)
[originally from svn r8396]
function. Useful in conjunction with the new --save option to
generate lots of games, extract the aux strings from the game
generator, and de-obfuscate them in order to measure statistical
properties of their solutions.
[originally from svn r8352]
generates PPC/Intel dual-architecture binaries.
This turns out not to be too painful: you compile and link your
programs using `gcc -arch ppc' or `gcc -arch i386', then you use a
command of the form `lipo -create ppc-binary i386-binary -output
binary' to construct a universal binary. It works equally well on
command-line standalone executable files and the executables within
application directories. Also added the -mmacosx-version-min option,
since otherwise the OS X build tools appear to default to building
binaries which will crash (without anything resembling a
comprehensible error message) on any earlier release.
The handling of version.o in this checkin is somewhat grotty. I'd
prefer a method more cleverly intertwingled with mkfiles.pl so I
didn't have to maintain the OS X architecture list in both
mkfiles.pl and Recipe. (Not that I anticipate Apple switching
architectures again in the immediate future, but it's the principle
of the thing.)
[originally from svn r7916]
is mostly done with ifdefs in windows.c; so mkfiles.pl generates a
new makefile (Makefile.wce) and Recipe enables it, but it's hardly
any different from Makefile.vc apart from a few definitions at the
top of the files.
Currently the PocketPC build is not enabled in the build script, but
with any luck I'll be able to do so reasonably soon.
[originally from svn r7337]
Windows puzzle binaries. This checkin involves several distinct
changes:
- mkfiles.pl now has an extra feature: if an object file is listed
in Recipe with a trailing question mark, it will be considered
optional, and silently dropped from the makefile if its primary
source file isn't present at the time mkfiles.pl runs. This means
people who check out the puzzles from Subversion and just run
mkfiles.pl shouldn't get build failures; they just won't get the
icons.
- all the .R files now use this feature to include an optional
Windows resource file.
- the .rc resource source files are built by icons/Makefile.
- windows.c finds the icon if present and uses it in place of the
standard Windows application icon.
[originally from svn r7020]
when testing a new game, so here's a new architecture for the Recipe
file. mkfiles.pl now supports several new features:
- an `!include' directive, which accepts wildcards
- += to append to an existing object group definition
- the ability to divert output to an arbitrary file.
So now each puzzle has a `.R' file containing a fragment of Recipe
code describing that puzzle, and the central Recipe does `!include
*.R' to construct the Makefiles. That way, I can keep as many
experimental half-finished puzzles lying around my working directory
as I like, and I won't have to keep reverting Recipe when I check in
any other changes.
As part of this change, list.c is no longer a version-controlled
file; it's now constructed by mkfiles.pl, so that it too can take
advantage of this mechanism.
[originally from svn r6781]
maxflow.c. Also in this checkin, fixes to the OS X and GTK back ends
to get ALIGN_VNORMAL right. This is the first time I've used it! :-)
[originally from svn r6390]
a non-recursive level above Easy, which therefore moves the
recursive Hard mode further up still. Play-testing suggests that in
fact Tricky is often _harder_ than the old Hard mode, since the
latter had limited depth of recursion and would therefore spot
complex deductions only if it happened to start a recursion on the
right square; Tricky may be limited in the sophistication of its
complex deductions, but it never misses one, so its puzzles tend to
be hard all over.
Also in this checkin, a new source file `nullfe.c', containing all
the annoying stub functions required to make command-line solvers
link successfully. James wrote this for (the new) lightupsolver, and
I've used it to simplify the other stand-alone solvers.
[originally from svn r6254]
it had a command-line solver. In order to do this, I've had to
expose the internal region numbering because the solver has to have
some way to state which region it means; and in any case it's also
useful to have human-visible region numbering so that two people can
discuss a puzzle they're solving together. So pressing L during play
now toggles the display of region numbers; and `mapsolver' uses
those same numbers when showing its working and its solutions.
[originally from svn r6244]
place in r6190. I'm quite pleased that I didn't have to modify the
printing infrastructure _at all_ to make this work; the only source
change required outside windows.c was the addition of a trivial
utility function midend_get_params(), and that was for the benefit
of bulk puzzle generation rather than anything to do with actual
printing.
As far as I can tell, all printable puzzles now print almost
indistinguishably from the way they print under Unix. If you look
closely the font is slightly different, and the Windows standard
hatching doesn't seem to be quite as nice as the kind I did by hand
in ps.c (and, particularly annoyingly, hatched areas don't show up
at all for me when I print to a file and use gv, though they come
out fine on the printer itself); but it's all there, and it all
works.
[originally from svn r6193]
[r6190 == af59dcf6858264103bbc621761feee3aed5aaf2a]
as seen by the back ends from the one implemented by the front end,
and shoved a piece of middleware (drawing.c) in between to permit
interchange of multiple kinds of the latter. I've also added a
number of functions to the drawing API to permit printing as well as
on-screen drawing, and retired print.py in favour of integrated
printing done by means of that API.
The immediate visible change is that print.py is dead, and each
puzzle now does its own printing: where you would previously have
typed `print.py solo 2x3', you now type `solo --print 2x3' and it
should work in much the same way.
Advantages of the new mechanism available right now:
- Map is now printable, because the new print function can make use
of the output from the existing game ID decoder rather than me
having to replicate all those fiddly algorithms in Python.
- the new print functions can cope with non-initial game states,
which means each puzzle supporting --print also supports
--with-solutions.
- there's also a --scale option permitting users to adjust the size
of the printed puzzles.
Advantages which will be available at some point:
- the new API should permit me to implement native printing
mechanisms on Windows and OS X.
[originally from svn r6190]
(This puzzle is theoretically printable, but I haven't added it in
print.py since there's rather a lot of painful processing required
to get from the game ID to the puzzle's visual appearance. It
probably won't become printable unless I get round to implementing a
more integrated printing architecture.)
[originally from svn r6186]
Also in this checkin (committed by mistake - I meant to do it
separately), a behind-the-scenes change to Slant to colour the two
non-touching classes of diagonals in different colours. Both colours
are set to black by default, but configuration by way of
SLANT_COLOUR_* can distinguish them if you want.
[originally from svn r6164]
nikoli.co.jp (which has quite a few puzzles that they don't seem to
have bothered to translate into English).
Minor structural change: the disjoint set forest code used in the
Net solver has come in handy again, so I've moved it out into its
own module dsf.c.
[originally from svn r6155]
generation) from a simple but rather fun Flash game I saw this
morning.
Small infrastructure change for this puzzle: while most game
backends find the midend's assumption that Solve moves are never
animated to be a convenience absolving them of having to handle the
special case themselves, this one actually needs Solve to be
animated. Rather than break that convenience for the other puzzles,
I've introduced a flag bit (which I've shoved in mouse_priorities
for the moment, shamefully without changing its name).
[originally from svn r6097]
command-line programs (solosolver, patternsolver, mineobfusc) to be
built as part of the normal Makefiles. This means mkfiles.pl now has
the capability to compile a source file more than once with
different #defines. Also, fixes for those auxiliary programs and one
fix in midend.c which the Borland compiler objected to while I was
testing its makefile generation.
[originally from svn r6066]
