Eliminate remaining gcc warnings

Various one-liners (mostly removing unused variables) to make the code
safe for the full set of warnings used by the xorg macros.

On Debian-based systems, flex generates incorrect code resulting in two
warnings about yy_getcolumn and yy_setcolumn having no previous
declaration despite being non-static.  Fedora carries a patch to fix
this, and a bug has been filed on Debian's flex to add the patch:
http://bugs.debian.org/667027

Aside from this, it's now safe for --enable-strict-compilation.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Don't parse nonsense Unicode/hex keycodes

If a keysym was specified as "U1039andsomeextrastuffontheend", return
NoSymbol rather than 0x10001039; similarly, return NoSymbol for
"0xdeadbeefhitherehowsyourdaybeen" rather than 0xdeadbeef.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove unused device_spec

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

include resets group compatibility modifiers #43091

This change makes sure that include does not overwrite previous
compatibility modifier settings when the included files does not
explicitly specify them.

Signed-off-by: Andreas Wettstein <wettstein509@solnet.ch>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
[Cross-picked from xkbcomp commit 14470719.]

Add cscope output to .gitignore

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Properly document xkb_key_get_syms

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Constify a global table

Signed-off-by: Ran Benita <ran234@gmail.com>

Make temporary fix for stale xkb_atoms

When xkb_free_keymap is called the atoms are all free'd, but action.c
keeps a global copy of interned "true" and "false", which remains stale.

The correct fix is to remove the need for the ActionsInit function
entirely.

Signed-off-by: Ran Benita <ran234@gmail.com>

Fix bugs to allow multiple keymaps in one process

These were several initializations that were forgotten in the previous
memory leak fixes.

Now several xkb_desc's can coexist (relatively) peacefully.

Signed-off-by: Ran Benita <ran234@gmail.com>
[daniels: Only the atom.c is relevant with the new context API.]

Remove useless check from xkb_intern_atom

The "makeit" variable is always true. Remove it and de-indent.
(Also change the type of the "len" variable to size_t to avoid some
useless casting).

Signed-off-by: Ran Benita <ran234@gmail.com>

Avoid leak in CompileKeymap error path

The NULL check is unneeded, and prevented the atoms from being free'd.

Signed-off-by: Ran Benita <ran234@gmail.com>
[daniels: Updated for xkb_map_unref.]

makekeys: use correct format strings

The new glibc (2.15) appear to cause trouble, particularly the sscanf
call, where makekeys will output empty hash tables. Using the appropriate
macros from inttypes.h makes it work again.

Signed-off-by: Ran Benita <ran234@gmail.com>

Add a NULL check before before strcmp'ing

The names array can have NULL entries for some virtual modifier indexes.

Signed-off-by: Ran Benita <ran234@gmail.com>

makekeys: free strings before exiting

Just to make valgrind happy.

Signed-off-by: Ran Benita <ran234@gmail.com>

Use straight strtok_r instead of wrapper

strtok_r is POSIX-2001 and should be fine.

Signed-off-by: Ran Benita <ran234@gmail.com>

Avoid use of partly initialized VModInfo in error path

Such as:

Compiling path: ./test/data/bad.xkb  mapName:
==1300== Conditional jump or move depends on uninitialised value(s)
==1300==    at 0x4E46166: HandleVModDef (vmod.c:90)
==1300==    by 0x4E3FEC9: HandleKeyTypesFile (keytypes.c:1035)
==1300==    by 0x4E3FBE1: HandleIncludeKeyTypes.constprop.11 (keytypes.c:387)
==1300==    by 0x4E401DD: HandleKeyTypesFile (keytypes.c:1022)
==1300==    by 0x4E3FBE1: HandleIncludeKeyTypes.constprop.11 (keytypes.c:387)
==1300==    by 0x4E401DD: HandleKeyTypesFile (keytypes.c:1022)
==1300==    by 0x4E4026F: CompileKeyTypes (keytypes.c:1150)
==1300==    by 0x4E3DF9B: CompileKeymap (keymap.c:169)
==1300==    by 0x4E465E9: compile_keymap (xkbcomp.c:205)
==1300==    by 0x4E46BE4: xkb_compile_keymap_from_file (xkbcomp.c:290)
==1300==    by 0x400B37: test_file (filecomp.c:47)
==1300==    by 0x4008E3: main (filecomp.c:90)
==1300==  Uninitialised value was created by a stack allocation
==1300==    at 0x4E3FB3F: HandleIncludeKeyTypes.constprop.11 (keytypes.c:366)

Signed-off-by: Ran Benita <ran234@gmail.com>

Fail gracefully on failure to find component include

If we can't find the component of the include file we're looking for,
make sure we don't return success when we meant failure, segfault, or
spectacularly leak everything.

Tested with incorrect component includes for keycodes, compat, symbols,
and types.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Reported-by: David Herrmann <dh.herrmann@googlemail.com>

Tests: Release context on failure to build keymap

No practical effect since they exit(1) regardless, but it keeps valgrind
happy.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Full support for multiple keysyms per level

Which also involved moving the global symbol map to be per-key instead;
this should probably be split out into a separate commit.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Parsing support for multiple keysyms per entry

Instead of allowing only one keysym per level per group, do as the
external API indicates and allow multiples.  The existing syntax is:
    key <AD01> { [ q, Q ] };
where the new syntax is:
    key <AD01> { [ q, Q, { H, E, L, L, O },
                 { Y, E, S, space, T, H, I, S, space, I, S, space, D, O, G } };

to make the key in the extreme top left of the keyboard do pretty
surprising things in levels 3 and 4.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Don't leak strings in merge declarations

Introduced by the reentrant scanner commit.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Don't leak include_paths in context

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove unused rtrnValue declaration

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Make parser and scanner reentrant

All global state is removed from the parser and scanner.
This makes use of the standard facilities in Bison and Flex for
reentrant/pure scanner/lexer and location tracking.

Signed-off-by: Ran Benita <ran234@gmail.com>
[daniels: Updated to current sources.]

Use xkb_contexts in keymap compilation

Primarily for the include path, but also for the logging in future.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Add new context API

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Update xkb_filter_group_lock_func for xkb_key_direction

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Fix distcheck with non-recursive Makefiles

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Fix build for X11 modifier masks

Exposed by include path changes, oops.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Handle group lock actions

The spec is simple here, as this action has no effect on key releases.

Signed-off-by: Ran Benita <ran234@gmail.com>

Make build non-recursive

Unify all the different Makefile.am into a single short top level one
(the test/Makefile.am file is left intact though).

This makes the build system simpler to look and should encourage
unifying more currently-disparate code.

Some further motivation can be found in this page:
http://www.flameeyes.eu/autotools-mythbuster/automake/nonrecursive.html

Signed-off-by: Ran Benita <ran234@gmail.com>

Define our own NoSymbol value and use it

Since we have our own xkb_keysym_t type, it makes sense to have our own
NoSymbol value instead of the one from X11/X.h.

Signed-off-by: Ran Benita <ran234@gmail.com>

Define our own None atom value

Since we define our own xkb_atom_t type, it makes sense not to use the
X11/X.h None value. This way we can also remove a lot of X11 includes.

Signed-off-by: Ran Benita <ran234@gmail.com>

Fix all -Wsign-compare warnings

i.e comparison of signed and unsigned values. These are mostly
harmless but fixing them allows to compile cleanly with -Wextra.

Signed-off-by: Ran Benita <ran234@gmail.com>

Silence -Wcast-qual warnings

There are some cases where we must free a string with a const qualifier.
Add a macro UNCONSTIFY to trick the compiler into silencing the warning
in the cases where we know what we're doing.

Signed-off-by: Ran Benita <ran234@gmail.com>

Use strcasecmp consistently instead of uStrCaseCmp

There's no use calling the same thing by a different name.

Signed-off-by: Ran Benita <ran234@gmail.com>

Move utility macro from XKBcommonint.h to utils.h

And merge all the similar ones into the same name.
The u* prefix is chosen over the _Xkb prefix because it has more uses
throughout the codebase. But It should now be simple to choose a nice
prefix and stay consistent.

Signed-off-by: Ran Benita <ran234@gmail.com>
[daniels: fixed for the case where we have strcasecmp]

Remove some more useless stuff

Some unused defines and geometry-removal leftovers (specifically the
file geom.c and the struct for the keyboard coordinates).

Signed-off-by: Ran Benita <ran234@gmail.com>

Rewrite recalloc to the correct type

The recalloc function should be expressed in terms of bytes to match its
name. However uTypedRecalloc retains its type so nothing is changed.

Signed-off-by: Ran Benita <ran234@gmail.com>

Remove useless stuff from utils

Signed-off-by: Ran Benita <ran234@gmail.com>
[daniels: fixed conflicts from strcasecmp, added includes to make
          filecomp build again]

Remove fallback strcasecmp/strncasecmp

Sorry if your libc doesn't have this, but it's not my problem.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Reported-by: Ran Benita <ran234@gmail.com>

Add include/ to makekeys include path

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Move utils.{c,h} to be used by the entire project

This is a first step for making consistent use of utils.h also outside
of xkbcomp/ .

Signed-off-by: Ran Benita <ran234@gmail.com>

Fix a scanf type mismatch

"%o" takes an unsigned int.

Signed-off-by: Ran Benita <ran234@gmail.com>

Remove a few remaining uses of the KeySym type

Signed-off-by: Ran Benita <ran234@gmail.com>

Add xkb test file to EXTRA_DIST

To appease make distcheck.

Signed-off-by: Ran Benita <ran234@gmail.com>

Fix incorrect usage of && instead of &

Signed-off-by: Ran Benita <ran234@gmail.com>

Update gitignore for 'state' test

Signed-off-by: Ran Benita <ran234@gmail.com>

Fix latin8 upper/lower case check

GCC complained:
misc.c:235:13: warning: logical 'and' of mutually exclusive tests is always false [-Wlogical-op]
misc.c:248:13: warning: logical 'and' of mutually exclusive tests is always false [-Wlogical-op]

Looking at keysymdef.h shows that the check is indeed wrong. This commit
updates the check to match the latin8 section of keysymdef exactly.

Signed-off-by: Ran Benita <ran234@gmail.com>

Properly free xkb_state's

Signed-off-by: Ran Benita <ran234@gmail.com>

Mention xkb_state_new can return NULL

in the header comments.

Signed-off-by: Ran Benita <ran234@gmail.com>

Move doxygen comment before enum item

Signed-off-by: Guillem Jover <guillem@hadrons.org>

Remove internal API from xkbcommon.h

And move it to XKBcomminint.h.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove hardcoded legacy modifier masks

Use the xkb_state_mod_* and xkb_map_mod_* API instead.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove unused changes structs

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove unused members of xkb_state

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Add enum xkb_key_direction instead of bool

Use XKB_KEY_UP instead of 0 and XKB_KEY_DOWN instead of 1.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Reported-by: Ran Benita <ran234@gmail.com>

Add state serialisation API

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Cosmetic coding style fixups

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Rename keymap allocation API

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Add LED state API

And also convert state.c to use the state API for mods and groups,
rather than testing the state members directly.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

LEDs: Fix string comparison bug

This resulted in duplicate sets of modifiers, since we were comparing
pointer equality of two strings, rather than string equality.  Oops.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Check for invalid indices in mod/group state API

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Fix documentation bugs with mod/group state API

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Add a proper README

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Round out new state API

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Fix (correct, but harmless) const warning

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Add new state API

Add new API to deal with xkb_state objects, including
xkb_state_update_key, which runs the XKB action machinery internally to
calculate what exactly happens to the state when a given key is pressed
or released.

The canonical way to deal with keys is now:
    struct xkb_state *state = xkb_state_new(xkb);
    xkb_keysym_t *syms;
    int num_syms;

    xkb_state_update_key(state, key, is_down);
    num_syms = xkb_key_get_syms(state, key, &syms);

More state handling API, including a way to get at or ignore preserved
modifiers, is on its way.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove some rogue fprintfs

Oops.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Don't clear real mods in action modmask

Previously, we would clear out the real modmask when updating the
modmask for action maps, if not using the key's modmask.  The correct
behaviour here is instead to use the key's modmask if using the modmap,
else use the real mods provided with the action originally.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove unnecessary allocation in expr.c

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Small constification

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Also print group name in keysym error

Unfortunately we can't get the actual file it was defined in this far
down, but at least give the human-readable name rather than just a group
index.

Also, groups are not zero-indexed, such that index 0 is group 1; fix
that too.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

More useful errors for invalid keysyms

Instead of generating a fairly droll internal error, generate a warning
also telling us exactly where the bad definition was.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Fix non-useModMapMods action masks

By ensuring their mask is only the vmods, rather than also potentially
including the key's modmap.  Also remove the unnecessary vmodmask
indirection.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Add geometry test

Add the dump of my full current X11/XKB keymap as a test for filecomp,
being as it also includes geometry.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Plug leaks in geometry parsing

When parsing a geometry file, don't leak the elements we've created.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove more float support

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Also update vmod -> indicator maps

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Update modifiers after building keymap

The server used to have to go and do this on our own, but we can do
better than that: after we've compiled the keymap, go through and bind
virtual modifiers to everything that needs it.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Convert interp action from xkb_any_action to xkb_action

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Add xkb_get_syms_by_level

Which allows you to pass in a group and a level rather than a
fully-formed state.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Fix leak in xkbcomp.c

Accidentally regressed this one briefly, oops.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove component listing code

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove unused control members

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Add a tiny bit of documentation to xkbcommon.h

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Add xkb_key_get_syms API

(And a slight cosmetic header reformatting.)

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Revert "Use XKB_COMMON_* modifier defines in ExprResolveModMask"

Because indices and masks are not at all the same thing.

This reverts commit 645275406f47369c9a67d02173aedf89e9d2a33c.

Use XKB_COMMON_* modifier defines in ExprResolveModMask

Rather than the ones from XKB.h.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove overlay support as well

Right now we just silently ignore overlay controls, which is probably
bad, but it's not the easiest to fix.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove half-implemented radio groups

It looks like this could never have worked anyway, what with num_rg
always being 0 everywhere.  Remove it.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove indirect XkbCM* and XkbSM* macros

The XkbKey* macros used to work through XkbCM* and XkbSM* variants, for
the xkb_map and xkb_server_map respectively; the latter versions weren't
used by anyone at all, so just bin them and make the macros work
directly.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Handle XkbFiles without a name

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Move allocation of xkb_desc into CompileKeymap

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Clean up xkb_keysym_to_string message

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove geometry even harder

Not the most elegant fix, but will do for now.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Introduce xkb_atom_t type

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Introduce xkb_keysym_t type

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Remove atom functions from public API

They're no longer needed since we don't expose any atoms in the
published API anymore.  As a result, we don't need to support external
atom implementations either.  Result!

Signed-off-by: Daniel Stone <daniel@fooishbar.org>

Stringify public name types

Ensure that all names under xkb_desc are strings, rather than atoms.

Signed-off-by: Daniel Stone <daniel@fooishbar.org>