chiark / gitweb /
sgt-puzzles.git
18 years agoRevert the code that assumed that incoming parameters used a random generation
Jacob Nevins [Tue, 5 Jul 2005 21:40:29 +0000 (21:40 +0000)]
Revert the code that assumed that incoming parameters used a random generation
scheme in r6068, since this messed up the non-permanency of the non-limited
parameter.

[originally from svn r6069]
[r6068 == f3c95109c7b905f74b5f149eea21f68444c31ef7]

18 years agoAdd a `full' parameter to validate_params(), analogous to the one in
Jacob Nevins [Tue, 5 Jul 2005 21:27:19 +0000 (21:27 +0000)]
Add a `full' parameter to validate_params(), analogous to the one in
encode_params(). This is necessary for cases where generation-time parameters
that are normally omitted from descriptive IDs can place restrictions on other
parameters; in particular, when the default value of a relevant generation-time
parameter is not the one used to generate the descriptive ID, validation could
reject self-generated IDs (e.g., Net `5x2w:56182ae7c2', and some cases in
`Pegs').

[originally from svn r6068]

18 years agoEnhancements to mkfiles.pl and Recipe to arrange for the auxiliary
Simon Tatham [Tue, 5 Jul 2005 19:40:32 +0000 (19:40 +0000)]
Enhancements to mkfiles.pl and Recipe to arrange for the auxiliary
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]

18 years agoIt's been a new year for a while now.
Jacob Nevins [Tue, 5 Jul 2005 19:38:27 +0000 (19:38 +0000)]
It's been a new year for a while now.

[originally from svn r6065]

18 years agorevert supposed typo -- I have no brain
Jacob Nevins [Tue, 5 Jul 2005 19:12:56 +0000 (19:12 +0000)]
revert supposed typo -- I have no brain

[originally from svn r6064]

18 years agoAdd HACKING to main doc build.
Jacob Nevins [Tue, 5 Jul 2005 19:11:51 +0000 (19:11 +0000)]
Add HACKING to main doc build.

[originally from svn r6063]

18 years agotypo
Jacob Nevins [Tue, 5 Jul 2005 19:10:26 +0000 (19:10 +0000)]
typo

[originally from svn r6062]

18 years agoAdd a completion flash when you get down to a single peg.
Jacob Nevins [Tue, 5 Jul 2005 18:57:34 +0000 (18:57 +0000)]
Add a completion flash when you get down to a single peg.

[originally from svn r6061]

18 years agoRefactored the game_size() interface, which was getting really
Simon Tatham [Tue, 5 Jul 2005 18:13:31 +0000 (18:13 +0000)]
Refactored the game_size() interface, which was getting really
unpleasant and requiring lots of special cases to be taken care of
by every single game. The new interface exposes an integer `tile
size' or `scale' parameter to the midend and provides two much
simpler routines: one which computes the pixel window size given a
game_params and a tile size, and one which is given a tile size and
must set up a drawstate appropriately. All the rest of the
complexity is handled in the midend, mostly by binary search, so
grubby special cases only have to be dealt with once.

[originally from svn r6059]

18 years agoJames Harvey points out a missing ifdef.
Simon Tatham [Tue, 5 Jul 2005 12:03:56 +0000 (12:03 +0000)]
James Harvey points out a missing ifdef.

[originally from svn r6058]

18 years agoClarify that diagonal moves are not permitted in Pegs.
Jacob Nevins [Mon, 4 Jul 2005 21:07:40 +0000 (21:07 +0000)]
Clarify that diagonal moves are not permitted in Pegs.

[originally from svn r6057]

18 years agoAdd WinHelp topic.
Jacob Nevins [Mon, 4 Jul 2005 20:06:35 +0000 (20:06 +0000)]
Add WinHelp topic.

[originally from svn r6056]

18 years agoThere's always one. Forgot to add Pegs to `make install'.
Simon Tatham [Mon, 4 Jul 2005 19:55:10 +0000 (19:55 +0000)]
There's always one. Forgot to add Pegs to `make install'.

[originally from svn r6054]

18 years agoDocumentation for Pegs.
Simon Tatham [Mon, 4 Jul 2005 19:53:36 +0000 (19:53 +0000)]
Documentation for Pegs.

[originally from svn r6053]

18 years agoPeg Solitaire implementation, complete with a random board
Simon Tatham [Mon, 4 Jul 2005 19:42:55 +0000 (19:42 +0000)]
Peg Solitaire implementation, complete with a random board
generator. The generator is lacking in almost any kind of finesse,
but it produces puzzles which at least _I_ find plausibly puzzling.

[originally from svn r6052]

18 years agoDocument keyboard controls for Flip
Jacob Nevins [Mon, 4 Jul 2005 14:35:14 +0000 (14:35 +0000)]
Document keyboard controls for Flip

[originally from svn r6051]

18 years agoFix bounds checking for Flip cursor control.
Jacob Nevins [Mon, 4 Jul 2005 11:21:50 +0000 (11:21 +0000)]
Fix bounds checking for Flip cursor control.

[originally from svn r6050]

18 years agoJames H's patch for a keyboard-controlled cursor in Flip.
Simon Tatham [Sun, 3 Jul 2005 14:49:00 +0000 (14:49 +0000)]
James H's patch for a keyboard-controlled cursor in Flip.

[originally from svn r6049]

18 years agoAllow game backends to use even special keystrokes such as N and Q;
Simon Tatham [Sun, 3 Jul 2005 11:45:49 +0000 (11:45 +0000)]
Allow game backends to use even special keystrokes such as N and Q;
they will only be processed as special by the midend if unwanted by
the backend. This causes 5x5 Solo to become just about playable,
because you can now click in a square and type `n'. However, typing
`n' when a square is not selected will revert to the normal
behaviour of starting a new game.

(This isn't particularly ideal, I know, but it's better than
nothing.)

[originally from svn r6048]

18 years agodraw_polygon() and draw_circle() have always had a portability
Simon Tatham [Sun, 3 Jul 2005 09:35:29 +0000 (09:35 +0000)]
draw_polygon() and draw_circle() have always had a portability
constraint: because some front ends interpret `draw filled shape' to
mean `including its boundary' while others interpret it to mean `not
including its boundary' (and X seems to vacillate between the two
opinions as it moves around the shape!), you MUST NOT draw a filled
shape only. You can fill in one colour and outline in another, you
can fill or outline in the same colour, or you can just outline, but
just filling is a no-no.

This leads to a _lot_ of double calls to these functions, so I've
changed the interface. draw_circle() and draw_polygon() now each
take two colour arguments, a fill colour (which can be -1 for none)
and an outline colour (which must be valid). This should simplify
code in the game back ends, while also reducing the possibility for
coding error.

[originally from svn r6047]

18 years agoJames Harvey points out that entering an invalid game ID can affect
Simon Tatham [Fri, 1 Jul 2005 16:50:49 +0000 (16:50 +0000)]
James Harvey points out that entering an invalid game ID can affect
the current midend state even if you don't subsequently enter a
valid one. Reorganise midend_game_id_int() so that (just like
midend_deserialise()) it does all its error checking before altering
anything in the midend's persistent data, so that it either succeeds
completely or fails before doing anything at all.

[originally from svn r6045]

18 years agoPrevent drags from being started outside the playing area, which was
Simon Tatham [Thu, 30 Jun 2005 18:11:41 +0000 (18:11 +0000)]
Prevent drags from being started outside the playing area, which was
causing invalid move descriptions to be returned from
interpret_move() and then failing an assertion when execute_move()
refused them.

[originally from svn r6044]

18 years agoPreset names retrieved from the environment must be dupstr()ed. How
Simon Tatham [Thu, 30 Jun 2005 18:11:02 +0000 (18:11 +0000)]
Preset names retrieved from the environment must be dupstr()ed. How
did I miss this before? It just caused a segfault for me, which is
entirely fair enough, but I've no idea why it didn't fail before!

[originally from svn r6043]

18 years agoLoad and Save are now supported on all three desktop platforms, and
Simon Tatham [Thu, 30 Jun 2005 18:00:37 +0000 (18:00 +0000)]
Load and Save are now supported on all three desktop platforms, and
documented. (This means the GTK temporary dependency on an
environment variable is now gone.)

[originally from svn r6042]

18 years agoGeneral robustness patch from James Harvey:
Simon Tatham [Thu, 30 Jun 2005 09:07:00 +0000 (09:07 +0000)]
General robustness patch from James Harvey:
 - most game_size() functions now work in doubles internally and
   round to nearest, meaning that they have less tendency to try to
   alter a size they returned happily from a previous call
 - couple of fiddly fixes (memory leaks, precautionary casts in
   printf argument lists)
 - midend_deserialise() now constructs an appropriate drawstate,
   which I can't think how I overlooked myself since I _thought_ I
   went through the entire midend structure field by field!

[originally from svn r6041]

18 years agoAttempting to begin a drag from (i.e. just click on) the hint pegs
Simon Tatham [Thu, 30 Jun 2005 08:25:47 +0000 (08:25 +0000)]
Attempting to begin a drag from (i.e. just click on) the hint pegs
for a past guess caused strangeness up to and including segfault,
thanks to bad bounds checking. Well spotted John Sullivan.

[originally from svn r6040]

18 years agoAdd a cast whose absence was causing a (-Werror-exacerbated) compile
Simon Tatham [Wed, 29 Jun 2005 16:32:53 +0000 (16:32 +0000)]
Add a cast whose absence was causing a (-Werror-exacerbated) compile
warning on OS X.

[originally from svn r6037]

18 years agoValidation of random-state-type game descriptions was broken. This
Simon Tatham [Wed, 29 Jun 2005 12:19:08 +0000 (12:19 +0000)]
Validation of random-state-type game descriptions was broken. This
meant that a Mines game saved before the first click had taken place
could not be successfully reloaded.

[originally from svn r6036]

18 years agoWhen serialising this game, it's probably better to keep the
Simon Tatham [Wed, 29 Jun 2005 08:20:49 +0000 (08:20 +0000)]
When serialising this game, it's probably better to keep the
half-finished guess in the game_ui rather than throwing it away.

[originally from svn r6033]

18 years agoMake peg removal accessible from the keyboard.
Jacob Nevins [Tue, 28 Jun 2005 18:10:20 +0000 (18:10 +0000)]
Make peg removal accessible from the keyboard.

[originally from svn r6032]

18 years agoNew {en,de}code_ui functions should be static. Oops.
Simon Tatham [Tue, 28 Jun 2005 17:43:50 +0000 (17:43 +0000)]
New {en,de}code_ui functions should be static. Oops.

[originally from svn r6031]

18 years agoActually implemented the serialise/deserialise functions in
Simon Tatham [Tue, 28 Jun 2005 17:05:05 +0000 (17:05 +0000)]
Actually implemented the serialise/deserialise functions in
midend.c. Also I've added an experimental front end in gtk.c only:
`Save' and `Load' options on the Game menu, which don't even show up
unless you define the magic environment variable
PUZZLES_EXPERIMENTAL_SAVE. Once I'm reasonably confident that the
whole edifice is plausibly stable, I'll take that out and turn it
into a supported feature (and also implement it in OS X and Windows
and write documentation).

[originally from svn r6030]

18 years agoMore serialisation changes: the game_aux_info structure has now been
Simon Tatham [Tue, 28 Jun 2005 11:14:09 +0000 (11:14 +0000)]
More serialisation changes: the game_aux_info structure has now been
retired, and replaced with a simple string. Most of the games which
use it simply encode the string in the same way that the Solve move
will also be encoded, i.e. solve_game() simply returns
dupstr(aux_info). Again, this is a better approach than writing
separate game_aux_info serialise/deserialise functions because doing
it this way is self-testing (the strings are created and parsed
during the course of any Solve operation at all).

[originally from svn r6029]

18 years agoRogue diagnostic!
Simon Tatham [Tue, 28 Jun 2005 08:35:55 +0000 (08:35 +0000)]
Rogue diagnostic!

[originally from svn r6028]

18 years agoMove a rogue declaration to the top of its block.
Simon Tatham [Tue, 28 Jun 2005 08:06:56 +0000 (08:06 +0000)]
Move a rogue declaration to the top of its block.

[originally from svn r6027]

18 years agoAnother function pair required for serialisation; these ones save
Simon Tatham [Tue, 28 Jun 2005 07:33:49 +0000 (07:33 +0000)]
Another function pair required for serialisation; these ones save
and restore anything vitally important in the game_ui. Most of the
game_ui is expected to be stuff about cursor positions and currently
active mouse drags, so it absolutely _doesn't_ want to be preserved
over a serialisation; but one or two things would be disorienting or
outright wrong to reset, such as the Net origin position and the
Mines death counter.

[originally from svn r6026]

18 years agoAnnoying special cases for Mines.
Simon Tatham [Tue, 28 Jun 2005 06:59:27 +0000 (06:59 +0000)]
Annoying special cases for Mines.

Firstly, the `Restart' function now reconstructs an initial game
state from the game description rather than dup_game(states[0]).
This means that Restart in a game of Mines restarts to just _after_
the initial click, so you can resume the puzzle-solving part without
having to remember where you placed that click.

Secondly, the midend now contains a second `private' game desc,
which is guaranteed to actually reconstruct the initial game_state
correctly (which Mines's publicly visible game descs tend not to,
since they describe a state which has already had the first click).
This should make serialising of Mines more sensible.

[originally from svn r6025]

18 years agoRe-architecting of the game backend interface. make_move() has been
Simon Tatham [Mon, 27 Jun 2005 19:34:54 +0000 (19:34 +0000)]
Re-architecting of the game backend interface. make_move() has been
split into two functions. The first, interpret_move(), takes all the
arguments that make_move() used to get and may have the usual side
effects of modifying the game_ui, but instead of returning a
modified game_state it instead returns a string description of the
move to be made. This string description is then passed to a second
function, execute_move(), together with an input game_state, which
is responsible for actually producing the new state. (solve_game()
also returns a string to be passed to execute_move().)

The point of this is to work towards being able to serialise the
whole of a game midend into a byte stream such as a disk file, which
will eventually support save and load functions in the desktop
puzzles, as well as restoring half-finished games after a quit and
restart in James Harvey's Palm port. Making each game supply a
convert-to-string function for its game_state format would have been
an unreliable way to do this, since those functions would not have
been used in normal play, so they'd only have been tested when you
actually tried to save and load - a recipe for latent bugs if ever I
heard one. This way, you won't even be able to _make_ a move if
execute_move() doesn't work properly, which means that if you can
play a game at all I can have pretty high confidence that
serialising it will work first time.

This is only the groundwork; there will be more checkins to come on
this theme. But the major upheaval should now be done, and as far as
I can tell everything's still working normally.

[originally from svn r6024]

18 years agoFix GTK casts to restore correct compilation on GTK 2.0 after r6022.
Simon Tatham [Sun, 26 Jun 2005 13:12:34 +0000 (13:12 +0000)]
Fix GTK casts to restore correct compilation on GTK 2.0 after r6022.

[originally from svn r6023]
[r6022 == e1c84c9a4ea9e74c128c28a77ea08336350991e3]

18 years agoUnder Gtk 1.2 (at least on Debian woody), a config or preset change that would
Jacob Nevins [Sun, 26 Jun 2005 12:54:08 +0000 (12:54 +0000)]
Under Gtk 1.2 (at least on Debian woody), a config or preset change that would
otherwise not cause the window size to change caused it to become very small
indeed. This change from Simon fixes that behaviour; I haven't tested it with
Gtk 2.

[originally from svn r6022]

18 years agoAdd debugging support.
Jacob Nevins [Sat, 25 Jun 2005 17:24:03 +0000 (17:24 +0000)]
Add debugging support.

[originally from svn r6018]

18 years agoFurther additions to the Guess docs.
Jacob Nevins [Sat, 25 Jun 2005 13:52:52 +0000 (13:52 +0000)]
Further additions to the Guess docs.

[originally from svn r6014]

18 years agoMake the keyboard-control cursors visible whenever a keyboard control is
Jacob Nevins [Sat, 25 Jun 2005 13:43:45 +0000 (13:43 +0000)]
Make the keyboard-control cursors visible whenever a keyboard control is
activated.

[originally from svn r6013]

18 years agoThe AngleArc() function that was being used to draw circles on Windows turns
Jacob Nevins [Sat, 25 Jun 2005 13:24:19 +0000 (13:24 +0000)]
The AngleArc() function that was being used to draw circles on Windows turns
out to be unsupported on the Win9x/Me series. Use Arc() instead (tested on
Win98 and Win2K).

[originally from svn r6012]

18 years agoCouple of presets.
Simon Tatham [Fri, 24 Jun 2005 16:35:42 +0000 (16:35 +0000)]
Couple of presets.

[originally from svn r6011]

18 years agoMore patches from James Harvey: enforce a maximum brightness in the
Simon Tatham [Fri, 24 Jun 2005 16:35:27 +0000 (16:35 +0000)]
More patches from James Harvey: enforce a maximum brightness in the
background colour (to ensure white pegs show up against it), and
convert the keyboard-control cursor into a rectangle when it's over
the hint pegs (otherwise it looks rather silly for numbers of pegs
above 4).

[originally from svn r6010]

18 years agoFix another display/UI glitch which triggered if you filled a row
Simon Tatham [Fri, 24 Jun 2005 12:30:30 +0000 (12:30 +0000)]
Fix another display/UI glitch which triggered if you filled a row
(so that the hint pegs lit up as `ready') and then pressed Undo: the
markable flag would remain set and the redrawing wouldn't darken the
pegs again.

[originally from svn r6009]

18 years agoTurn off the keyboard-control cursor when the user reverts to mouse
Simon Tatham [Fri, 24 Jun 2005 11:13:08 +0000 (11:13 +0000)]
Turn off the keyboard-control cursor when the user reverts to mouse
control.

[originally from svn r6008]

18 years agoChange the preprocessor symbol `DEBUG' to `DEBUGGING', since the
Simon Tatham [Fri, 24 Jun 2005 11:05:43 +0000 (11:05 +0000)]
Change the preprocessor symbol `DEBUG' to `DEBUGGING', since the
former is automatically defined by Cygwin.

[originally from svn r6007]

18 years agoMy background-erasing changes in r5996/r5997 caused trouble with the
Simon Tatham [Fri, 24 Jun 2005 11:03:20 +0000 (11:03 +0000)]
My background-erasing changes in r5996/r5997 caused trouble with the
keyboard-control cursors, and when I tried to fix those more
wallpaper bubbles popped up elsewhere. Here's what I think is a
proper fix: a comprehensive overhaul of the redraw code such that,
instead of tracking the cursor positions explicitly in the
drawstate, we instead track for each peg position whether or not a
cursor is currently displayed at that position. So cursor erasing
and cursor drawing become part of the main draw loop rather than a
separate bit on the end.

[originally from svn r6006]
[r5996 == 3d58feb561ef09cf977f710d69a3562529e23f0f]
[r5997 == a46e3266558eaf5cd18ac2a6322af65c196c3655]

18 years agoArray overflow fix from James Harvey.
Simon Tatham [Thu, 23 Jun 2005 23:11:59 +0000 (23:11 +0000)]
Array overflow fix from James Harvey.

[originally from svn r6005]

18 years agoMake the `hold marker' in Guess accessible from the keyboard (`H' key, for want
Jacob Nevins [Thu, 23 Jun 2005 22:01:17 +0000 (22:01 +0000)]
Make the `hold marker' in Guess accessible from the keyboard (`H' key, for want
of a better idea).

[originally from svn r6004]

18 years agoFactual and other corrections/additions to the Guess docs.
Jacob Nevins [Thu, 23 Jun 2005 21:47:55 +0000 (21:47 +0000)]
Factual and other corrections/additions to the Guess docs.

[originally from svn r6003]

18 years agoReinstate WinHelp topic now we have a manual chapter.
Jacob Nevins [Thu, 23 Jun 2005 21:37:05 +0000 (21:37 +0000)]
Reinstate WinHelp topic now we have a manual chapter.

[originally from svn r6002]

18 years agoCredit James Harvey.
Simon Tatham [Thu, 23 Jun 2005 19:01:01 +0000 (19:01 +0000)]
Credit James Harvey.

[originally from svn r6000]

18 years agoJames Harvey's extensions to Guess: a couple of extra game settings
Simon Tatham [Thu, 23 Jun 2005 18:50:58 +0000 (18:50 +0000)]
James Harvey's extensions to Guess: a couple of extra game settings
plus a manual chapter.

[originally from svn r5999]

18 years agoAnd forgot to add guess to the Unix make install target! I need a list.
Simon Tatham [Thu, 23 Jun 2005 18:05:19 +0000 (18:05 +0000)]
And forgot to add guess to the Unix make install target! I need a list.

[originally from svn r5998]

18 years agoBah, not good enough. Extend those background erases by one more
Simon Tatham [Thu, 23 Jun 2005 18:02:21 +0000 (18:02 +0000)]
Bah, not good enough. Extend those background erases by one more
pixel.

[originally from svn r5997]

18 years agoErase backgrounds when drawing over an existing circle, for the
Simon Tatham [Thu, 23 Jun 2005 17:43:43 +0000 (17:43 +0000)]
Erase backgrounds when drawing over an existing circle, for the
benefit of antialiasing platforms such as OS X. Also in this
checkin, fiddle with svn:ignore (there's a new puzzle binary).

[originally from svn r5996]

18 years agoAllow dragging of coloured pegs from previous guesses. Also
Simon Tatham [Thu, 23 Jun 2005 16:36:09 +0000 (16:36 +0000)]
Allow dragging of coloured pegs from previous guesses. Also
reorganise the colours so there are fewer of those terribly
computery cyan and magenta shades, and more good old-fashioned
colours with simple names like orange and purple. Finally, change
the `right place' marking peg colour from red to black, in line with
at least _my_ old Mastermind set (I faintly suspect red marker pegs
of being an Americanism) and also so that the marker pegs and the
coloured pegs have no colours in common.

[originally from svn r5995]

18 years agoMust outline as well as filling circles, because not all platforms
Simon Tatham [Thu, 23 Jun 2005 12:55:47 +0000 (12:55 +0000)]
Must outline as well as filling circles, because not all platforms
can be guaranteed to draw bare filled shapes the same way.

[originally from svn r5994]

18 years agoTwo small fixes: add some ceil() calls to stop the puzzle shrinking
Simon Tatham [Thu, 23 Jun 2005 12:21:14 +0000 (12:21 +0000)]
Two small fixes: add some ceil() calls to stop the puzzle shrinking
every time I press N, and revert the help topic to NULL until some
documentation materialises.

[originally from svn r5993]

18 years ago`Guess', a Mastermind clone from James Harvey. This checkin also
Simon Tatham [Thu, 23 Jun 2005 09:14:19 +0000 (09:14 +0000)]
`Guess', a Mastermind clone from James Harvey. This checkin also
introduces a few new utility functions in misc.c, one of which is
the bitmap obfuscator from Mines (which has therefore been moved out
of mines.c).

[originally from svn r5992]

18 years agoIntroduce a front-end function to draw circles.
Simon Tatham [Thu, 23 Jun 2005 08:24:52 +0000 (08:24 +0000)]
Introduce a front-end function to draw circles.

[originally from svn r5991]

18 years agoThis TODO comment should have been taken out in r5913 :-)
Simon Tatham [Wed, 22 Jun 2005 09:26:03 +0000 (09:26 +0000)]
This TODO comment should have been taken out in r5913 :-)

[originally from svn r5988]
[r5913 == 02035753f817173a6861d1fc4bec437508cec42d]

18 years agoNew front end functions to save and restore a region of the puzzle
Simon Tatham [Wed, 22 Jun 2005 08:30:31 +0000 (08:30 +0000)]
New front end functions to save and restore a region of the puzzle
bitmap. Can be used to implement sprite-like animations: for
example, useful for games that wish to implement a user interface
which involves dragging an object around the playing area.

[originally from svn r5987]

18 years agoAdd WinHelp topic
Jacob Nevins [Tue, 21 Jun 2005 11:16:36 +0000 (11:16 +0000)]
Add WinHelp topic

[originally from svn r5984]

18 years agoPatch from James Harvey in response to the new Rectangles grid
Simon Tatham [Tue, 21 Jun 2005 08:15:53 +0000 (08:15 +0000)]
Patch from James Harvey in response to the new Rectangles grid
generator: (a) take out the `#ifdef SLOW_SYSTEM' because the new
generator is much more efficient, and (b) convert another int to
size_t to prevent the solver going negative in 16-bit ints as a
result.

[originally from svn r5983]

18 years agoConversation with Richard and Chris yesterday gave rise to a more
Simon Tatham [Mon, 20 Jun 2005 17:32:45 +0000 (17:32 +0000)]
Conversation with Richard and Chris yesterday gave rise to a more
sensible means of generating an initial gridful of rectangles. This
was previously a stupidly non-scalable bit of the Rectangles puzzle
generator: it filled a ludicrously large array with every possible
rectangle that could go anywhere in the grid, picked one at random
and winnowed the list by removing anything that overlapped that one,
then repeated until the list was empty (and therefore the grid was
full except for remaining singleton squares). Total cost was O(N^4)
in both time and space; not pretty.

Richard and Chris's sensible alternative was to place each rectangle
by randomly choosing a so-far-uncovered _square_, and then picking a
random rectangle from the possible ones covering that square. This
means we only have to deal with a small fragment of the rectangle
list at any one time, and we don't have to store the whole lot in
memory; so it's _much_ faster and more scalable, and has virtually
no memory cost.

A side effect of this algorithmic change is that the probability
distribution has altered. When you line up all the possible
_rectangles_ and pick one at random, then obviously the small ones
are going to be in the majority since lots of small ones can fit
into the space taken up by any given big one. So the original
algorithm tends to favour fiddly grids full of lots of tiny
rectangles, which don't tend to be very interesting. But if you
first pick a square and then think about the rectangles that can
surround that square, the small ones are suddenly going to be in the
_minority_ because there are only two ways you can place (say) a 2x1
containing a given square compared to 36 ways you can place a 6x6.
So this algorithm favours more large rectangles, which I generally
consider to be an improvement.

[originally from svn r5982]

18 years agoJames Harvey's memory leak patch for Flip.
Simon Tatham [Sun, 19 Jun 2005 21:49:53 +0000 (21:49 +0000)]
James Harvey's memory leak patch for Flip.

[originally from svn r5980]

18 years agoRedraw glitch: tiles marked black (at game-over time) were not
Simon Tatham [Sun, 19 Jun 2005 14:33:06 +0000 (14:33 +0000)]
Redraw glitch: tiles marked black (at game-over time) were not
redrawn as non-black on undo. Introduce a new flag TILE_IMPOSSIBLE,
so that information about those black markers is cached in the
drawstate and we know when we have to erase them.

In the process I've removed the game_state argument completely from
the subfunction tile_redraw(), which gives me some confidence that
it isn't getting any _more_ privileged information out of it.

[originally from svn r5979]

18 years agoJust noticed yesterday that initial window sizing is broken on
Simon Tatham [Sat, 18 Jun 2005 08:52:50 +0000 (08:52 +0000)]
Just noticed yesterday that initial window sizing is broken on
Windows for puzzles with status bars, because the initial call to
check_window_size is given the window size _without_ the status bar
and assumes that that has to be big enough for the whole thing
_with_ the status bar, so it shrinks everything by a little bit. So
now we resize the window to take account of the status bar before
calling check_window_size(), and the problem seems to have gone away.

[originally from svn r5975]

18 years agoI've decided I didn't like the asymmetry of putting the solution
Simon Tatham [Sat, 18 Jun 2005 08:46:33 +0000 (08:46 +0000)]
I've decided I didn't like the asymmetry of putting the solution
markers in the top left of the square.

[originally from svn r5974]

18 years agoI'm sick of not having a `make install' target.
Simon Tatham [Sat, 18 Jun 2005 08:36:52 +0000 (08:36 +0000)]
I'm sick of not having a `make install' target.

[originally from svn r5973]

18 years agoCorrect rogue chapter into a subheading.
Simon Tatham [Fri, 17 Jun 2005 19:06:25 +0000 (19:06 +0000)]
Correct rogue chapter into a subheading.

[originally from svn r5972]

18 years agoSolver for Flip.
Simon Tatham [Fri, 17 Jun 2005 18:55:36 +0000 (18:55 +0000)]
Solver for Flip.

[originally from svn r5970]

18 years agoInfrastructure change which I've been thinking about for a while:
Simon Tatham [Fri, 17 Jun 2005 18:54:58 +0000 (18:54 +0000)]
Infrastructure change which I've been thinking about for a while:
the back end function solve_game() now takes the _current_
game_state in addition to the initial one.

[originally from svn r5969]

18 years agoAnimation.
Simon Tatham [Fri, 17 Jun 2005 17:32:28 +0000 (17:32 +0000)]
Animation.

[originally from svn r5968]

18 years agoAnother new puzzle! This one isn't particularly deep or complex
Simon Tatham [Fri, 17 Jun 2005 17:16:49 +0000 (17:16 +0000)]
Another new puzzle! This one isn't particularly deep or complex
(solving it only requires matrix inversion over GF(2), whereas
several of the other puzzles in this collection are NP-complete in
principle), but it's a fun enough thing to play with and is
non-trivial to do in your head - especially on the hardest preset.

[originally from svn r5967]

18 years agoAn email conversation with Chuck Fresno turned up several forms of
Simon Tatham [Fri, 17 Jun 2005 11:51:52 +0000 (11:51 +0000)]
An email conversation with Chuck Fresno turned up several forms of
symmetry which were not implemented in Solo. Now they are.

In the process I've completely retired symmetry_limit() on the
grounds that some of the new symmetries do not have a rectangular
base region; instead I determine the base region by going through
the grid and finding every square which is not transformed into a
lexicographically lower square by any symmetry operation. This means
that adding new symmetries is now _only_ a matter of encoding the
actual transformation rules.

[originally from svn r5965]

18 years agoClean up the behaviour of drags off the edge of the grid. These have
Simon Tatham [Fri, 10 Jun 2005 13:05:24 +0000 (13:05 +0000)]
Clean up the behaviour of drags off the edge of the grid. These have
always been a convenient way to cancel a drag you decided was an
error, but now it's more obvious that this is the case.

[originally from svn r5941]

18 years agoDisable shuffle overlap checking in the special case w=h=n.
Simon Tatham [Fri, 10 Jun 2005 11:34:02 +0000 (11:34 +0000)]
Disable shuffle overlap checking in the special case w=h=n.

[originally from svn r5940]

18 years agoOverzealous last-move tracking during shuffle was causing lockups on
Simon Tatham [Fri, 10 Jun 2005 11:29:59 +0000 (11:29 +0000)]
Overzealous last-move tracking during shuffle was causing lockups on
2x2 grid generation.

[originally from svn r5939]

18 years agoPatch from James Harvey to rearrange the Same Game colours.
Simon Tatham [Fri, 10 Jun 2005 11:27:26 +0000 (11:27 +0000)]
Patch from James Harvey to rearrange the Same Game colours.

[originally from svn r5938]

18 years agoPatch from Chris Emerson to add a status bar to Rectangles. Apart
Simon Tatham [Fri, 10 Jun 2005 11:24:44 +0000 (11:24 +0000)]
Patch from Chris Emerson to add a status bar to Rectangles. Apart
from saying `Auto-solved' or `Completed', the main function of the
status bar is to track the current size of a rectangle you're
dragging out, which makes life much easier when trying to count up
large rectangles in 19x19 mode.

[originally from svn r5937]

18 years agoGDK 2.6, in its doubtless infinite wisdom, has decided that
Simon Tatham [Thu, 9 Jun 2005 18:29:11 +0000 (18:29 +0000)]
GDK 2.6, in its doubtless infinite wisdom, has decided that
gdk_font_from_description() will no longer even _try_ to return a
font matching the specified Pango font description; instead it will
return `fixed' no matter what you do. Therefore, I've had to switch
to using Pango proper for Puzzles text rendering, rather than just
using Pango for font selection.

[originally from svn r5936]

18 years agoRemove debug output.
Jacob Nevins [Wed, 8 Jun 2005 09:20:35 +0000 (09:20 +0000)]
Remove debug output.

[originally from svn r5926]

18 years agoAdd Windows help topic
Jacob Nevins [Tue, 7 Jun 2005 21:05:29 +0000 (21:05 +0000)]
Add Windows help topic

[originally from svn r5923]

18 years agoTweaks and more complete documentation for Same Game.
Jacob Nevins [Tue, 7 Jun 2005 21:03:14 +0000 (21:03 +0000)]
Tweaks and more complete documentation for Same Game.

[originally from svn r5922]

18 years agoInteger overflow in game_size(). Oops.
Simon Tatham [Tue, 7 Jun 2005 20:44:14 +0000 (20:44 +0000)]
Integer overflow in game_size(). Oops.

[originally from svn r5921]

18 years agoChris Emerson points out that gtk_window_resize() isn't in GTK 1.2.
Simon Tatham [Tue, 7 Jun 2005 20:25:25 +0000 (20:25 +0000)]
Chris Emerson points out that gtk_window_resize() isn't in GTK 1.2.
Work around it the same way I did in PuTTY.

[originally from svn r5920]

18 years agoJames Harvey's patch to support keyboard control in Same Game.
Simon Tatham [Tue, 7 Jun 2005 20:22:25 +0000 (20:22 +0000)]
James Harvey's patch to support keyboard control in Same Game.

[originally from svn r5919]

18 years agoForgot to add Same Game to the big list for OS X.
Simon Tatham [Tue, 7 Jun 2005 20:22:08 +0000 (20:22 +0000)]
Forgot to add Same Game to the big list for OS X.

[originally from svn r5918]

18 years agoJames Harvey (again) points out an array underrun in the new
Simon Tatham [Tue, 7 Jun 2005 19:07:58 +0000 (19:07 +0000)]
James Harvey (again) points out an array underrun in the new
clash-checking code in Solo. (valgrind confirms it.)

[originally from svn r5916]

18 years agoAnother James Harvey patch. This one introduces a new button code
Simon Tatham [Tue, 7 Jun 2005 19:04:52 +0000 (19:04 +0000)]
Another James Harvey patch. This one introduces a new button code
called `CURSOR_SELECT', intended for platforms which have a fire
button to go with their cursor keys. (Apparently some Palms do.)
CURSOR_SELECT is not bound to anything on any of the current three
platforms, and therefore no game may rely on it being available.
This checkin adds it in Net as a synonym for 'a', the keyboard-based
rotate-left command.

[originally from svn r5915]

18 years agoJames Harvey has contributed an implementation of `Same Game', also
Simon Tatham [Tue, 7 Jun 2005 19:01:36 +0000 (19:01 +0000)]
James Harvey has contributed an implementation of `Same Game', also
known as ksame (KDE) and Same GNOME (GNOME).

[originally from svn r5914]

18 years agoAll the games in this collection have always defined their graphics
Simon Tatham [Tue, 7 Jun 2005 17:57:50 +0000 (17:57 +0000)]
All the games in this collection have always defined their graphics
in terms of a constant TILE_SIZE (or equivalent). Here's a
surprisingly small patch which switches this constant into a
run-time variable.

The only observable behaviour change should be on Windows, which
physically does not permit the creation of windows larger than the
screen; if you try to create a puzzle (Net makes this plausible)
large enough to encounter this restriction, the Windows front end
should automatically re-adjust the puzzle's tile size so that it
does fit within the available space.

On GTK, I haven't done this, on the grounds that X _does_ permit
windows larger than the screen, and many X window managers already
provide the means to navigate around such a window. Gareth said he'd
rather navigate around a huge Net window than have it shrunk to fit
on one screen. I'm uncertain that this makes sense for all puzzles -
Pattern in particular strikes me as something that might be better
off shrunk to fit - so I may have to change policy later or make it
configurable.

On OS X, I also haven't done automatic shrinkage to fit on one
screen, largely because I didn't have the courage to address the
question of multiple monitors and what that means for the entire
concept :-)

[originally from svn r5913]

18 years agoIntroduce a new game backend function (there seem to have been a lot
Simon Tatham [Mon, 6 Jun 2005 11:21:36 +0000 (11:21 +0000)]
Introduce a new game backend function (there seem to have been a lot
of these recently) whose job is to update a game_ui to be consistent
with a new game_state. This is called by midend.c in every situation
where the current game_state changes _other_ than as a result of
make_move (Undo, Redo, Restart, Solve).

The introduction of this function allows a game_ui to contain
information about selections or highlights within a game_state which
simply wouldn't make sense when transferred to another game_state.
In particular, I've used it to fix a subtle bug in Solo whereby,
although you couldn't right-click to pencil-mode highlight a filled
square, you could _get_ a pencil-mode highlight in a filled square
if you used Undo and Redo. (Undo to before the square was filled,
right-click to highlight it, then Redo. Alternatively, left-click
and clear the square, right-click to highlight it, then Undo.)

[originally from svn r5912]

18 years agoSmall UI bug: LEFT_RELEASEs were being thrown away completely if
Simon Tatham [Sat, 4 Jun 2005 17:51:49 +0000 (17:51 +0000)]
Small UI bug: LEFT_RELEASEs were being thrown away completely if
their coordinates were outside the playing area. Clearly no actual
move should be made in that situation, but we do at least need to
sort out any highlighted squares from the prior dragging operations.

[originally from svn r5911]

18 years agoRevamp pencil mark placement and sizing. Pencil marks are now sized
Simon Tatham [Sat, 4 Jun 2005 12:32:40 +0000 (12:32 +0000)]
Revamp pencil mark placement and sizing. Pencil marks are now sized
and positioned according to how many there are in the cell, rather
than how many distinct digits there are in the entire puzzle. This
means that 4x4 Solo can now be played with pencil marks without
_too_ much difficulty; the marks will still get a bit crowded if you
have more than 12 in the same square, but with luck that shouldn't
happen often, and as long as you're down in the 2-9 range things
should be entirely legible.

[originally from svn r5909]