This is hit when passing an empty string and XKB_KEYSYM_CASE_INSENSITIVE
to xkb_keysym_from_name currently if `(lo + hi) / 2` is 0 and `cmp < 0`,
causing mid to underflow and the the array access into name_to_keysym on
the next iteration of the loop to be out of bounds .
We *would* use ssize_t here as it is the appropriate type, but windows
unfortunately does not define it.
In 7d84809fdc I added a fast path for the
case-sensitive case, but it is still slowing down Compose parsing.
Instead of the binary search, use a perfect hash function, computed with
a simple python module I found (vendored).
It is faster -- perf diff is:
Baseline Delta Abs Shared Object Symbol
........ ......... ................. ...................................
22.35% -14.04% libc-2.33.so [.] __strcmp_avx2
16.75% +10.28% bench-compose [.] xkb_keysym_from_name
20.72% +2.40% bench-compose [.] parse.constprop.0
2.29% -1.97% bench-compose [.] strcmp@plt
2.56% +1.81% bench-compose [.] resolve_name
2.37% +0.92% libc-2.33.so [.] __GI_____strtoull_l_internal
26.19% -0.63% bench-compose [.] lex
1.45% +0.56% libc-2.33.so [.] __memchr_avx2
1.13% -0.31% libc-2.33.so [.] __strcpy_avx2
Also reduces the binary size:
Before:
text data bss dec hex filename
341111 5064 8 346183 54847 build/libxkbcommon.so.0.0.0
After:
text data bss dec hex filename
330215 5064 8 335287 51db7 build/libxkbcommon.so.0.0.0
Note however that it's still larger than before 7d84809fdccbb5898d08388:
text data bss dec hex filename
320617 5168 8 325793 4f8a1 build/libxkbcommon.so.0.0.0
Signed-off-by: Ran Benita <ran@unusedvar.com>
Relatedly, strtoul allows a lot of unwanted stuff (spaces, +/- sign,
thousand seperators), we really ought not use it. But that's for another
time.
Signed-off-by: Ran Benita <ran@unusedvar.com>
xkb_keysym_from_name() is called a lot in Compose file parsing. The
lower case handling slows things down a lot (particularly given we can't
use the optimized strcasecmp() due to locale issues). So add separate
handling for the non-case-sensitive case which is used by Compose.
To do this we need to add another version of the ks_tables table. This
adds ~20kb to the shared library binary. We can probably do something
better here but I think it's fine.
Signed-off-by: Ran Benita <ran@unusedvar.com>
Previously we included it with an `-include` compiler directive. But
that's not portable. And it's better to be explicit anyway.
Every .c file should have `include "config.h"` first thing.
Signed-off-by: Ran Benita <ran@unusedvar.com>
This avoids the problem that MSVC does not provide strcasecmp() nor
strncasecmp(), and at the same time avoids potential problems due to
locale configuration by using istrcmp() and istrncmp() which are
already in the source tree and written to cover only ASCII.
lowercase: LATIN SMALL LETTER SHARP S (U+00DF)
uppercase: LATIN CAPITAL LETTER SHARP S (U+1E9E)
The uppercase sharp s (XK_ssharp) is a relatively recent addition to unicode
but was added to the relevant keyboard layouts in xkeyboard-config-2.25
(d1411e5e95c)
https://gitlab.freedesktop.org/xkeyboard-config/xkeyboard-config/issues/144
Alas, the CapsLock behavior was broken on the finnish layout (maybe others).
This was due XConvertCase() never returning the uppercase characters.
Let's make this function return the right lower/upper symbols for the sharp s
and hope that the world won't get any worse because of it.
Corresponding Xlib issue:
https://gitlab.freedesktop.org/xorg/lib/libx11/issues/110
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
These can be useful in some odd cases.
There is already an implementation (+ tests) for internal use, so all
that's needed is to export them.
If xkbcommon were to provide a way to convert a Unicode codepoint to a
keysym, this could have been implemented externally as follows:
uint32_t codepoint = xkb_keysym_to_utf32(keysym);
uint32_t upper_codepoint = my_unicode_library_to_upper(codepoint);
xkb_keysym_t upper_keysym = theoretical_xkb_keysym_from_utf32(upper_codepoint);
However keysym -> codepoint is not injective so such a function is not
possible strictly speaking.
Signed-off-by: Ran Benita <ran234@gmail.com>
We currently use strcasecmp, which is locale-dependent. In particular,
one well-known surprise even if restricted just ASCII input is found in
the tr_TR (Turkish) locale, see e.g.
https://msdn.microsoft.com/en-us/library/ms973919.aspx#stringsinnet20_topic5
We have known to avoid locale-dependent functions before, but in this
case, we forgot.
Fix it by implementing our own simple ASCII-only strcasecmp/strncasecmp.
Might have been possible to use strcasecmp_l() with the C locale, but
went the easy route.
Side advantage is that even this non-optimized version is faster than
the optimized libc one (__strcasecmp_l_sse42) since it doesn't need to
do the locale stuff. xkb_keysym_from_name(), which uses strcasecmp
heavily, becomes faster, and so for example Compose file parsing, which
uses xkb_keysym_from_name() heavily, becomes ~20% faster.
Resolves https://github.com/xkbcommon/libxkbcommon/issues/42
Signed-off-by: Ran Benita <ran234@gmail.com>
Instead of thinking about signed <-> unsigned an whatnot.
bsearch() is inline in glibc, so gcc optimizes this away anyway.
Signed-off-by: Ran Benita <ran234@gmail.com>
This adds a flags argument to xkb_keysym_from_name() so we can perform a
case-insensitive search. This should really be supported as many keysyms
have really weird capitalization-rules.
However, as this may produce conflicts, users must be warned to only use
this for fallback paths or error-recovery. This is also the reason why the
internal XKB parsers still use the case-sensitive search.
This also adds some test-cases so the expected results are really
produced. The binary-size does _not_ change with this patch. However,
case-sensitive search may be slightly slower with this patch. But this is
barely measurable.
[ran: use bool instead of int for icase, add a recommendation to the
doc, and test a couple "thorny" cases.]
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
This removes the complicated and undocumented hash-table creation-helper
and replaces it with an autogenerated sorted array. The search uses simple
bsearch() now.
We also tried using gperf but it turned out to generate way to big
hashtables and when reducing the size it isn't really faster than
bsearch() anymore.
There are no users complaining about the speed of keysym lookups and we
have no benchmarks that tell that we are horribly slow. Hence, we can
safely use the simpler approach and drop all that old code.
Signed-off-by: Ran Benita <ran234@gmail.com>
Signed-off-by: David Herrmann <dh.herrmann@googlemail.com>
Our current code (taken from the xserver) doesn't handle unicode keysyms
at all, and there seem to be some other changes compared to libX11,
which is what xkbcomp uses. So we just copy the code that does that from
libX11.
It would be much better to not have to hardcode unicode tables like
that, but it's probably better than dealing with glibc locale stuff for
now. It also doesn't affect our binary size much.
Signed-off-by: Ran Benita <ran234@gmail.com>
This is useful to see whether the function was successful and whether
truncation occurred.
It just changes void -> int so shouldn't break API or ABI.
Signed-off-by: Ran Benita <ran234@gmail.com>
- Add context.h and move context-related functions from xkb-priv.h to
it.
- Move xkb_context definition back to context.c.
- Add keysym.h and move keysym upper/lower/keypad from xkb-priv.h to it.
- Rename xkb-priv.h to map.h since it only contains keymap-related
definitions and declarations now.
- Remove unnecessary includes and some and some other small cleanups.
Signed-off-by: Ran Benita <ran234@gmail.com>
With Dan Nicholson's permission (via email), update his copyright and
license statements to the standard X.Org boilerplate MIT license, as
both myself and Ran have been using.
Clean up my copyright declarations (in some cases to correct ownership),
and add copyright/license statements from myself and/or Ran where
appropriate.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
It's more tidy and less error prone, since we use strcasecmp == 0 a lot.
We replace strcmp == 0 by streq, strcasecmp == 0 by istreq,
uStrCasePrefix by istreq_prefix and uDupString by strdup_safe.
Signed-off-by: Ran Benita <ran234@gmail.com>
.uncrustify.cfg committed for future reference also, but had to manually
fix up a few things: it really likes justifying struct initialisers.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
This avoids a couple of special cases in the code, and is more
consistent. Since anyone who includes xkbcommon.h also gets
xkbcommon-keysyms.h, and anyone who include xkbcommon-keysyms.h would
want NoSymbol anyway, there's no down side.
Signed-off-by: Ran Benita <ran234@gmail.com>
Change them to refer to the string representation of the keysym's name
as a name rather than a string, since we want to add API to get the
Unicode printable representation as well.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Make the files in the src/* directory use their own header or a
consilidated private header. This makes the file dependencies clearer.
Also drop the pointless "xkb" file name prefix, add split a few
declarations to their own files (atom.h and text.h).
Signed-off-by: Ran Benita <ran234@gmail.com>
The definitions in config.h should be available in all files an
implementation detail; it can be included through the build system
instead of having each file pull it every time.
This is especially helpful with AC_USE_SYSTEM_EXTENSIONS, as _GNU_SOURCE
and friends can have an effect by merely being defined, which can lead
to some confusion if its effective for only half the files.
And we don't really support a build _without_ config.h; so, one less
thing to worry about.
Signed-off-by: Ran Benita <ran234@gmail.com>
The kbproto header is already not needed here anymore.
Move the _X_EXPORT's to the corresponding function definitions, and use
straight extern "C" clauses instead of _XFUNCPROTOBEGIN/END.
It also makes more sense to have the EXPORT's in the source files, as it
provides some documentation to the reader, whereas in the header it's
obvious.
Signed-off-by: Ran Benita <ran234@gmail.com>
[daniels: Updated for xkb_keymap changes.]
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>
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>
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>
This integrates two commits from libX11:
ebd6ef0a4db0ddef0ae17ad14571518ccdeea5ba
XStringToKeysym: Special case for XF86 keysyms
Some XFree86 keysyms were in XKeysymDB as XF86_foo, despite really being
XF86foo. So, if we get to the bottom of XStringToKeysym and haven't
found our XF86_foo, try it again as XF86foo.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
00175397480b76d32bf82b0c7c94c91a2a95954e
makekeys: Scan vendor keysyms as well as core
Since we can't really live without vendor keysyms, scan them all in to
generate ks_tables.h, rather than only doing the core ones, and leaving
the vendor syms to be manually synchronised with XKeysymDB.
Signed-off-by: Daniel Stone <daniel@fooishbar.org>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Notice that the xkey.sh test is changed to match libX11 behavior, i.e.
XKeysymToString(0x1008FE20) -> "XF86Ungrab" as opposed to "XF86_Ungrab".
Signed-off-by: Ran Benita <ran234@gmail.com>
This matches commit 24283d40b1e4314c6647dda49d2a159833341a8b from
libX11:
Since makekeys is built using build environment's compiler and
runs natively, we have to make sure that the size of the
Signature type is the same on both the native environment
and the target, otherwise we get mismatches upon running X,
and some LSB test failures (xts5).
Use an unsigned 32-bit integer on all platforms.
Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
Reviewed-by: Daniel Stone <daniel@fooishbar.org>
Signed-off-by: Ran Benita <ran234@gmail.com>
Isaac Freund
Ran Benita
Adrian Perez de Castro
Peter Hutterer
Ran Benita
Jasper St. Pierre
David Herrmann
Daniel Stone
Kristian Høgsberg