Since these lines are always orthogonal, it's easier to draw them
using draw_rect than draw_thick_line.
Previously, the grid lines were drawn just inside the top and left
edges of the region redrawn by draw_square(), so only the bottom and
right edges of the whole grid were not covered by any draw_square
call. To avoid having to shift the logical grid centre, I'm now
drawing parts of the grid outline on all four sides of the
draw_square() region, so that half the thickened grid edge protrudes
on every side of the grid as a whole.
In the process of splitting up the grid line width into the part on
the top and left edges and the part on the bottom and right, I've
renamed the confusingly named BORDER_WIDTH define (which wasn't used
anyway) to a set of things that make it clear that they're referring
to the grid lines as opposed to the border of the whole puzzle.
The default of 1/30 is rather thin, and probably wasn't chosen
deliberately (since it was just inherited from the default 1-pixel
line width, and the preferred tile size).
Thicker crosses stand out more and make play easier.
Use 1/16 since it's a rounder number than 1/15 :-).
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
Simply replace the calls to draw_line with calls to draw_thick_line.
We need to choose a thickness parameter. The preferred tile size is
30, and the "draw_line" function draws a 1-pixel line, so the
thickness right now is 1/30 the tile size, at the preferred size.
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
A line less than 1 pixel wide may not be visible. So if a backend
wants to draw a line whose width scaled by the window size, that line
thickness ought to be at least 1.0.
That way if the scale is small, but still big enough that there is a
straightforward interpretation of the drawing primitives which is
legible, we implement that interpretation.
If a frontend draws a narrower line, making it wider might cause
drawing anomalies, due to the line now having a bigger bounding box.
These anomalies should occur only at small scales where currently the
display is not legible, and we should fix them as we notice them.
Signed-off-by: Ian Jackson <ijackson@chiark.greenend.org.uk>
gcc objects to this in -pedantic mode, which means other compilers are
technically entitled to object too if they like. Usually I try to
avoid it by putting a dummy value at the end of the enum, but I forgot
in this case.
(And I didn't notice, because _my_ local builds run without -pedantic,
on the grounds that configure.ac autodetects that my system's GTK
headers are not pedantic-clean. Oh well.)
Now, with an odd grid size, we choose the posterisation threshold so
that half the time it delivers ceil(n/2) black squares and half the
time it delivers floor(n/2). Previously it only did the former, which
meant that asking Pattern to generate a 1x1 puzzle (with the bug in
the previous commit fixed) would always generate the one with a single
black square, and never the one with a single white square. Both are
trivial to solve, of course, but it seemed inelegant!
No change to the number of black squares in the puzzle solution can
constitute a spoiler for the player, of course, because that number is
trivial to determine without doing any difficult reasoning, just by
adding up all the clues in one dimension.
We were filling in a row immediately as all-white if it had no clues
at all, but weren't filling in a row as all-black if it had a single
clue covering the entire row. Now we do both.
In particular, this caused the Pattern solver to be unable to take
advantage of one of the two kinds of totally obvious clue across the
_easy_ dimension of a trivial 1xN puzzle - and a special case of
_that_, as a user pointed out, is that the game generator hangs trying
to create a 1x1 puzzle, which ought to be the easiest thing in the
world!
This should make it less annoying for me to do local testing of the JS
output of a build, before I push a change. There's a new
build.out/jstest directory containing .html files suitable for loading
in a local browser, which refer to the JS files via an appropriate
relative path to the existing build.out/js directory.
This is in addition to the existing keystrokes r, ^R and ^Y. I've
become used to Ctrl-Shift-Z in other GUI games, and my fingers keep
getting confused when my own puzzles don't handle it the same way.
This fixes an amusing UI bug that I think can currently only come up
in the unpublished puzzle 'Group', but there's no reason why other
puzzles _couldn't_ do the thing that triggers the bug, if they wanted
to.
Group has unusual keyboard handling, in that sometimes (when a cell is
selected for input and the key in question is valid for the current
puzzle size) the game's interpret_move function will eat keystrokes
like 'n' and 'u' that would otherwise trigger special UI events like
New Game or Undo.
The bug is that fake keypress events generated from the GUI menus
looked enough like those keystrokes that interpret_move would eat
those too. So if you start, say, a 16x16 Group puzzle, select an empty
cell, and then choose 'new game' from the menu, Group will enter 'n'
into the cell instead of starting a new game!
I've fixed this by inventing a new set of special keystroke values
called things like UI_NEWGAME and UI_UNDO, and having the GUI menus in
all my front ends generate those in place of 'n' and 'u'. So now the
midend can tell the difference between 'n' on the keyboard and New
Game from the menu, and so Group can treat them differently too. In
fact, out of sheer overcaution, midend.c will spot keystrokes in this
range and not even _pass_ them to the game back end, so Group
shouldn't be able to override these special events even by mistake.
One fiddly consequence is that in gtk.c I've had to rethink the menu
accelerator system. I was adding visible menu accelerators to a few
menu items, so that (for example) 'U' and 'R' showed up to the right
of Undo and Redo in the menu. Of course this had the side effect of
making them real functioning accelerators from GTK's point of view,
which activate the menu item in the same way as usual, causing it to
send whatever keystroke the menu item generates. In other words,
whenever I entered 'n' into a cell in a large Group game, this was the
route followed by even a normal 'n' originated from a real keystroke -
it activated the New Game menu item by mistake, which would then send
'n' by mistake instead of starting a new game!
Those mistakes cancelled each other out, but now I've fixed the
latter, I've had to fix the former too or else the GTK front end would
now undo all of this good work, by _always_ translating 'n' on the
keyboard to UI_NEWGAME, even if the puzzle would have wanted to treat
a real press of 'n' differently. So I've fixed _that_ in turn by
putting those menu accelerators in a GtkAccelGroup that is never
actually enabled on the main window, so the accelerator keys will be
displayed in the menu but not processed by GTK's keyboard handling.
(Also, while I was redoing this code, I've removed the logic in
add_menu_item_with_key that reverse-engineered an ASCII value into
Control and Shift modifiers plus a base key, because the only
arguments to that function were fixed at compile time anyway so it's
easier to just write the results of that conversion directly into the
call sites; and I've added the GTK_ACCEL_LOCKED flag, in recognition
of the fact that _because_ these accelerators are processed by a weird
mechanism, they cannot be dynamically reconfigured by users and
actually work afterwards.)
That's a case in which the current game IDs have changed, so the
midend ought to be calling the front-end function (if any) that
notifies it when that happens.
The only front end of mine that was affected by this missing call was
the Javascript one, which uses that callback to update the 'Link to
this puzzle' links below the game canvas - but, of course, that front
end didn't ever call midend_deserialise until this month, so no wonder
I never noticed before.
(But downstream front ends might be affected too, for all I know.)
Commit ef39e6e17 made a goof in which the 'New game' button had no
border on the left and an accidental extra one on the right, which I'm
really not sure how I failed to spot when I tested it yesterday.
The only user to send me a comment today on the new layout said that
that menu item in particular is annoying to have hidden behind more
clicks, so by a vote of one to nothing, it's back out in the open.
I've been thinking for a while that it's about time I did that. Those
images used to link to the Java versions of the puzzles, back when
Java was the in-browser applet platform of choice; then I made them
not link to either one when it wasn't clear which system would win;
but now JS has clearly won, it's past time the images linked to the JS
puzzles, as the obviously sensible default.
This is done by showing a dialog containing an <input type="file">
through which the user can 'upload' a save file - though, of course,
the 'upload' doesn't go to any HTTP server, but only into the mind of
the Javascript running in the same browser.
It would be even nicer to support drag-and-drop as an alternative UI
for getting the save file into the browser, but that isn't critical to
getting the first version of this feature out of the door.
This is done by getting midend_serialise to produce the complete
saved-game file as an in-memory string buffer, and then encoding that
into a data: URI which we provide to the user as a hyperlink in a
dialog box. The hyperlink has the 'download' attribute, which means
clicking on it should automatically offer to save the file, and also
lets me specify a not-too-silly default file name.
I'm about to want to call these from Javascript as well as from
Emscripten-compiled C, so I need versions of them that aren't wrapped
up in the Emscripten library object and also don't expect half their
parameters to be Emscripten-verse char pointers requiring calls to
Pointer_stringify.
The easiest way to achieve all of that is to turn the Emscripten-
ready versions of those functions in emcclib.js into tiny wrappers
around the JS versions, which do the pointer stringification and a
couple of other details like directing callbacks to the right C
functions.
I'm about to introduce a few more options, and the button bar is
already a bit wide, so I think I should shrink it horizontally before
putting more stuff on it. So I've organised the options into something
more like the Game and Type submenus that the desktop versions use.
However, I haven't gone quite all the way, on the basis that the web
versions will be at least slightly playable on devices without much
keyboard, which makes it worth keeping the in-play actions (Undo,
Redo, and to a lesser extent Restart and Solve) accessible as
top-level buttons in a single click each.
As part of this change, I've also separated the menu bar into a
drop-down menus section and a top-level buttons section with a gap
between them, and put a signalling "..." on the end of the titles in
the former section.
This change also removes the class="left" on the game-type menu and
its submenus, which were previously there to deal with that menu being
all the way over on the far right of the menu bar. But the CSS for
those classes is still there in jspage.pl, and should still work if I
need it again in future.
This is intended to make life easier for the _really_ dense grids in
which the generator algorithm falls back to a bare clearing, a tightly
packed section round the edges, and a fringe of deductions required in
between.
In that situation, you can deduce _in principle_ from the remaining-
mines counter that there are (say) one, or two, squares left to be
uncovered before everything remaining has to be a mine. And often the
game will require that deduction in order to solve it all by pure
logic. But actually doing it requires counting up the huge number of
covered squares in an irregularly shaped area and subtracting the mine
count in the status line, which is a real pain.
In fact, people failing to do that are the biggest source of (wrong)
bug reports about Mines games having no solution, so with any luck
this will make my own life easier.
This also switches them to being 64-bit, which I think is
probably acceptable in this modern age, especially for such
a non-essential piece of software. If anyone complains I can
always reinstate a parallel 32-bit build.
To support the switch to 64-bit, this commit also changes the default
install directory in the MSI to 'Program Files' rather than 'Program
Files (x86)'.
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.
I'm getting rid of these installers in general, and also I'm about to
switch the Windows builds over to my new-look non-Windows non-Wine
system, which can't run the Inno Setup builder anyway.
The solver code still had an assumption, which must have dated before
the Solve menu option was introduced, that all puzzles presented to it
had at least one valid solution, and was enforcing that assumption by
assert(). Now the solver returns a more sensible failure code which
solve_game() can convert into a proper error message.
Apparently new versions of GCC get grumpy if the possible range for a sprintf()
call exceeds MAXINT, which would never happen in actuality here due to the size
of the puzzles we're dealing with... but the compiler doesn't know that, of
course, so thinks that something may have gone horribly awry. Changing it to a
char solves the problem neatly.
Thanks to Glen Sawyer for reporting it. This is surely a consequence
of separating interpret_move from execute_move - if I'd done things in
the more obvious way, then this bug would never have happened, because
once the autofollow code had gone once round the loop it would find
that the starting edge was no longer in the same state it was looking
for. But since we don't start changing the states of edges until
execute_move(), it's possible for interpret_move() to go round and
round a cycle for ever.
Fortunately, it can _only_ happen if the edge you clicked on was part
of a loop which is the whole of its connected component - i.e. every
vertex in the cycle has degree 2 - and therefore we don't need O(N)
space to detect it (e.g. by recording whether each edge has been
visited already), but instead we can simply check if we've come back
to the starting edge.
We enforce by assertion that the target buffer size is nonzero before
subtracting 1 from it; the call to fatal() is replaced by another
assert so that it will give clearer diagnostic information if it
fails; the variable holding the return value of strlen should be
size_t and its declaration should be in a C90-compatible location.
Finally, the reason why the function needs to be exist is clarified.
Rockbox's sprintf() lacks the ability to left-justify a string. Fixed
by adding a copy_left_justfied() function to misc.c.
This is a new version of this commit, as the previous version broke
saving!
This is mostly intended to make play more convenient for grid types
like the new Great-Great-Dodecagonal, and other grids with very
high-degree faces, in which it's annoying to have to spend half a
dozen mouse clicks on filling in a path of edges round the outside of
one of those faces which clearly have to all be set (or clear) if any
one of them is.
For now, the new feature is enabled by yet another of my hacky
environment variables, called LOOPY_AUTOFOLLOW. You can set it to
"off", "fixed" or "adaptive", where "off" is currently the default
(hence, no user-visible change in the default behaviour from this
change). If set to 'fixed', then toggling the state of any edge will
automatically toggle any further edges which are in the same state and
share a degree-2 vertex of the underlying grid design with the
original one. In 'adaptive' mode, the game goes even further, and will
consider outgoing edges in LINE_NO state not to count for purposes of
deciding if a vertex has degree 2.
This is the game for which I bothered to introduce the feature at all.
Because of the large number of grid types, the presets menu was
getting quite unwieldy; but because the grid dimensions for each grid
type are more or less arbitrary, it's still useful to have at least
one reasonably sized example of each grid type. So I've compromised by
moving some of the grid types into a 'More' submenu.
(I'm not particularly wedded to _which_ settings deserve relegation. I
may change my mind and move things about later on.)
To do this, I've completely replaced the API between mid-end and front
end, so any downstream front end maintainers will have to do some
rewriting of their own (sorry). I've done the necessary work in all
five of the front ends I keep in-tree here - Windows, GTK, OS X,
Javascript/Emscripten, and Java/NestedVM - and I've done it in various
different styles (as each front end found most convenient), so that
should provide a variety of sample code to show downstreams how, if
they should need it.
I've left in the old puzzle back-end API function to return a flat
list of presets, so for the moment, all the puzzle backends are
unchanged apart from an extra null pointer appearing in their
top-level game structure. In a future commit I'll actually use the new
feature in a puzzle; perhaps in the further future it might make sense
to migrate all the puzzles to the new API and stop providing back ends
with two alternative ways of doing things, but this seemed like enough
upheaval for one day.
The previous control buttons and dropdowns based on form elements were
always a bit ugly: partly in a purely visual sense, and partly because
of the nasty bodge I had to do with splitting the usual 'Custom' game
type menu item into two (to get round the fact that if an element of a
<select> is already selected, browsers won't send an event when it's
re-selected). Also, I'm about to want to introduce hierarchical
submenus in the Type menu, and <select> doesn't support that at all.
So here's a replacement system which does everything by CSS
properties, including the popping-up of menus when the mouse moves
over their parent menu item. (Thanks to the Internet in general for
showing me how that trick is done.)
The use of plain numbers was needlessly confusing, and in particular
made it too easy to make unintended changes to the existing Loopy
presets when inserting a new grid enum value anywhere other than at
the end of the list.
But in the course of doing this I realised that, against all
sensibleness, the numeric indices for grid types in grid.h and in
Loopy itself don't match up! Right now I don't want to get sidetracked
into fixing the structural confusion that made that happen in the
first place, but I've at least materialised Loopy's version of the
enum with clearly identifiable LOOPY_GRID_* names.
Officially known as the '3-4-6-12 tiling', according to, e.g.,
https://en.wikipedia.org/wiki/3-4-6-12_tiling .
Thanks to Michael Quevillon for contributing this patch (and for
choosing a less hard-to-remember name for the tiling :-).
Another oddity involving an empty square is that if it coincides with
the source square for highlights (either by original design of the
game id, or because the player Ctrl-moves the source square into an
empty grid cell during play), then everything stops being lit up as
active. That's fine - you can still play the game using other
indications of error, such as the loop detection highlight - but it
looks silly for the status line to say 'Active: 1/lots'. So in that
situation I suppress the 'active' counter completely; it comes back
when you move the source square to somewhere it's _possible_ to
highlight more than one square.
While I'm at it, I've also removed the active counter in the case
where the game is completely solved, because in that situation it's
more or less unnecessary anyway, and that way the normal course of
play on the default small grid size doesn't overflow the available
status line space.
A hand-typed grid is permitted to use the square type '0' (never
generated by Net's own grid generator), which is a completely empty
square. This requires an adjustment to the completion checker, so that
such squares aren't required to be connected; otherwise, a grid
containing one would be permanently uncompletable.
However, the completion checker missed one case - it was
unconditionally checking that all squares are connected to the _top
left corner_, on the basis that (before I thought of the zero square)
any source square is as good as any other if what you really want to
know is whether they're all connected to each other. But that means
that if the top left square _is_ the empty one, things to wrong - e.g.
5x5:02c328ade11adb129d7c3e524 would fail to give a completion flash.
Fixed by starting the completion-checking search from the first
non-empty square we find.
gdk_window_set_background_rgba is deprecated as of GTK 3.22, because
apparently you can't just _say_ any more 'here is what I want my
window's background colour to be in places where a widget isn't'.
Instead you have to provide a GtkStyleProvider which can be slotted
into a wobbly tower of other providers with associated priorities, so
that the user can override your choices if they really want to.
And the easiest way to constructc a GtkStyleProvider in turn is to
write *actual CSS* and get GTK to parse it, so I end up converting my
nice numeric RGB values into a complicated text format for another
part of _the same process_ to parse back into numbers. Sigh.
I've just noticed that the GTK game window was not being redrawn when
changing between puzzle modes that don't involve a window resize - by
selecting a preset from the Type menu (e.g. changing between different
12x12 settings in Flood) or via the Custom menu.
It looks as if the bug was introduced in commit 8dfe5cec3, which
suggests that it was a side effect of the switch from
gtk_window_resize_to_geometry to plain gtk_window_resize. My guess is
that the implementation of the former function inside GTK might have
happened to trigger an unconditional window resize, while the latter
took the shortcut of doing nothing if the window was already the right
size; hence, resize_fe() would have been reliably generating a redraw
event without me having to ask for one, but now it doesn't, so I have
to trigger one myself any time I've just called resize_fe.
I was accidentally re-checking the value of component_state[comp]
after resetting comp to the special value -1, which caused a failure
to highlight stray path-shaped components if they existed in addition
to a large loop component. (Plus, of course, it's just illegal no
matter what visible behaviour it does or doesn't cause in practice.)
Fixed by adjusting the code to more closely match the version in Loopy
(I wonder how I managed to add two pieces of code in commit b31155b73
without noticing this difference between them).
A user points out that the error check that should have detected a
non-digit where a digit should have been was never firing, due to an
&& that should have been ||.
I don't think it was a harmful error - the subsequent check that the
number was in range, plus the skipping past only digits to find the
next part of the string, combine to arrange that not many kinds of
invalid game id could actually get through.
But it did have the small effect that a 0 could be elided without
triggering an error, e.g. the game ids
4:0,0,0,0,0,0L,0,0,0R,0U,0,0L,0,0,,3,
4:0,0,0,0,0,0L,0,0,0R,0U,0,0L,0,0,0,3,
would both be accepted, and would be decoded into the same game, even
though the former should have failed syntax validation. Now it does.
Fix errors pointed out by clang
error: logical not is only applied to the left hand side of this bitwise operator [-Werror,-Wlogical-not-parentheses]
| if (only_immutable && !copy->flags[i] & FLAG_IMMUTABLE) continue;
| ^
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Now we work out for ourselves how the drawing-area size relates to the
overall window size, by adding on the height of fe->menubar and/or
fe->statusbar.
Chris Boyle points out that two of the rules are implicitly intended
to be read as only applying if a previous rule hadn't already decided
what would happen, and suggested that since not all readers infer that
priority order, it would be better to explicitly make them mutually
exclusive so that there can be no confusion about which one applies.
The check_valid() function was not verifying that each Killer cage
summed to the right thing! Thanks to Chris Goodyer for spotting it. I
wonder how nobody else has, in 8 years.
In commits 24848706e and adc54741f, I revamped the highlighting of
erroneous connected components of those two puzzles' solution graphs
in cases where a non-solution loop existed, so that the largest
component was considered correct and the smaller ones lit up in red.
I intended this to work in the cases where you have most of a correct
solution as one component and a small spurious loop as another (in
which case the latter lights up red), or conversely where your mostly
correct component was joined into a loop leaving a few edges out (in
which case the left-out edges again light up red). However, a user
points out that I overlooked the case where your mostly correct
solution is not all one component! If you've got lots of separate
pieces of path, and one tiny loop that's definitely wrong, it's silly
to light up all but the longest piece of path as if they're erroneous.
Fixed by treating all the non-loop components as one unit for these
purposes. So if there is at least one loop and it isn't the only thing
on the board, then we _either_ light up all loops (if they're all
smaller than the set of non-loop paths put together), _or_ light up
everything but the largest loop (if that loop is the biggest thing on
the board).
It isn't necessary to cause the right files to _exist_, because
makedist.sh is run from Buildscr which has already run mkauto. But it
turns out it _is_ important to get the relative timestamps of
Makefile.in and Makefile.am the right way round, otherwise somebody
who unpacks the tarball and runs 'configure' and 'make' will find make
tries to rebuild Makefile.in because it thinks Makefile.am is newer -
and if they don't have the right automake installed, or any automake,
that will fail.
