This reverts commit d6d6cba9.

The reverted commit intended to fix the problem where an unpadded X
event struct is passed into XPutBackEvent, by creating a padded struct
with _XEventToWire and _XWireToEvent. However, _XWireToEvent updates the
last sequence number in Display, which may cause xlib to complain about
lost sequence numbers.

IMO, the problem that commit tried to solve is a bug in the client
library, and workaround it inside Xlib is bad practice, especially given
the problem it caused. Plus, the offender cited in the original commit
message, freeglut, has already fixed this problem.

Fixes: #176 #174

Signed-off-by: Yuxuan Shui <yshuiv7@gmail.com>

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>

Keith Packard authored 2 years ago and Jeremy Huddleston Sequoia committed 2 years ago

It seems to be common practice of some X11 clients to pass specific event
types into APIs that take XEvent*.  For example, freeglut does:

   XConfigureEvent fakeEvent = {0};
   ...
   XPutBackEvent(fgDisplay.Display, (XEvent*)&fakeEvent);

This can result in reads overflowing the input event when libX11 does:

   XEvent store = *event;

=================================================================
==75304==ERROR: AddressSanitizer: stack-buffer-overflow on address 0x00016ee4a8e8 at pc 0x000101c54d14 bp 0x00016ee4a0d0 sp 0x00016ee49888
READ of size 192 at 0x00016ee4a8e8 thread T0
    #0 0x101c54d10 in __asan_memcpy+0x1a4 (libclang_rt.asan_osx_dynamic.dylib:arm64e+0x3cd10)
    #1 0x102848a18 in _XPutBackEvent PutBEvent.c:41
    #2 0x1028490a4 in XPutBackEvent PutBEvent.c:84
    #3 0x1013295c8 in fgOpenWindow freeglut_window.c:1178
    #4 0x101321984 in fgCreateWindow freeglut_structure.c:108
    #5 0x10132b138 in glutCreateWindow freeglut_window.c:1551
    #6 0x100fb7d94 in main+0x78 (checkeredTriangles:arm64+0x100003d94)
    #7 0x197de3e4c  (<unknown module>)

Address 0x00016ee4a8e8 is located in stack of thread T0 at offset 840 in frame
    #0 0x1013282f8 in fgOpenWindow freeglut_window.c:1063

  This frame has 8 object(s):
    [32, 40) 'title.addr'
    [64, 176) 'winAttr' (line 1066)
    [208, 240) 'textProperty' (line 1067)
    [272, 352) 'sizeHints' (line 1068)
    [384, 440) 'wmHints' (line 1069)
    [480, 672) 'eventReturnBuffer' (line 1070)
    [736, 740) 'num_FBConfigs' (line 1072)
    [752, 840) 'fakeEvent' (line 1074) <== Memory access at offset 840 overflows this variable

This change allows XPutBackEvent() to support such clients without
risk of memory read overflow.

Reviewed-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>
Tested-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>

imLcPrs.c:681:52: runtime error: applying zero offset to null pointer

Signed-off-by: Jeremy Huddleston Sequoia <jeremyhu@apple.com>

Ulrich Sibiller authored 2 years ago and Alan Coopersmith committed 2 years ago

Is no longer a bool but a counter.

Signed-off-by: Ulrich Sibiller <uli42@gmx.de>

Ulrich Sibiller authored 2 years ago and Alan Coopersmith committed 2 years ago

Use the same indentation as the surrounding code.

Signed-off-by: Ulrich Sibiller <uli42@gmx.de>
Reviewed-by: Alan Coopersmith <alan.coopersmith@oracle.com>

Matthieu Herrb authored 2 years ago and Alan Coopersmith committed 2 years ago

- the activation logic is reversed
- there is also _XInternalLockDisplay() that needs protection
- I've found cases (in fvwm2) where the callback calls XCheckIfEvent()
  recursively. So the flag needs to be a counter.

Reviewed-by: Adam Jackson <ajax@redhat.com>

nia authored 2 years ago and Alan Coopersmith committed 2 years ago

XKBBind.c:230: error: #pragma GCC diagnostic not allowed inside functions

Signed-off-by: Thomas Klausner <wiz@gatalith.at>

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>

Po Lu authored 2 years ago and Alan Coopersmith committed 2 years ago

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>

Benno Schulenberg authored 2 years ago and Alan Coopersmith committed 2 years ago

It makes more sense to have similar sequences grouped together
than to rigidly follow the order of ascending Unicode codes.

Benno Schulenberg authored 2 years ago and Alan Coopersmith committed 2 years ago

The last few occurrences of `leftcaret` and `rightcaret` were replaced
with `less` and `greater` in xkeyboard-config half a year ago.

Benno Schulenberg authored 2 years ago and Alan Coopersmith committed 2 years ago

There was not yet any way to compose the Unicode minus sign, U+2212.

Adam Jackson authored 2 years ago and Adam Jackson committed 2 years ago

The documentation for this family of functions very clearly says not to
call into xlib in your predicate function, but historically single
threaded apps could get away with it just fine, and now that we've
forced thread-safety on the world such apps will now deadlock instead.
That's not an acceptable regression even if the app is technically
broken. This has been reported with XFCE and FVWM, and Motif's
cut-and-paste code has the same bug by inspection, so this does need to
be addressed.

This change nerfs LockDisplay/UnlockDisplay while inside the critical
bit of an IfEvent function. This is still safe in the sense that the
display remains locked and no other thread should be able to change it
from under us, but the loop that scans the event queue might not be
robust against it being modified as a side effect of protocol emitted by
the predicate callback. But that's not new, non-XInitThreads'd xlib
would have the same caveat.

Closes: #157

Several other `<Multi_key> <asciicircum> <symbol>` sequences
produce the superscript equivalent of the given symbol.  So,
let `<Multi_key> <asciicircum> <minus>` do the same.

Also, add two other sequences for producing a plain macron,
to compensate a bit the loss of the above sequence.

Additionally, make `<Multi_key> <underscore> <minus>` produce
a subscript minus, for consistency.

This fixes issue #165.

Requested-by: J. McWilliams

Signed-off-by: Benno Schulenberg <bensberg@telfort.nl>

Serbian used sr_YU (Yugoslavia) code in the past.
Then it was succeeded by sr_CS (Serbia and Montenegro).
Finally, it was split into sr_RS (Serbia) and sr_ME (Montenegro).

da95ecbb
introduced the modern sr_RS and sr_ME codes.

Next, 40761898
added the Serbian compose table but only for the legacy sr_CS entry.

5cd60398
removed the legacy sr_CS entry, the only one pointing to the correct compose file.
It also renamed the file to sr_RS, but did not update the compose mapping.

Let’s point all Serbian locales to the Compose file again.

Fixes: 701e9e9a ("Use same pthread-stubs as libxcb")

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>

Benno Schulenberg authored 2 years ago and Alan Coopersmith committed 2 years ago

Two years ago, commit b126bfd7 allowed using also a lowercase `u`
wherever an uppercase `U` was used to represent a breve.  But the
commit should have limited itself to only the prefixed uses of `U`,
as that is how most letters with a breve are composed.

Also, group the two compose sequences with an uppercase post-fixed `U`
together with the corresponding other post-fixed sequences.

Signed-off-by: Benno Schulenberg <bensberg@telfort.nl>

Avoid conflicts when libX11 calls libxcb and gets its pthread functions
overriding our ancient stubs.

v2: Keep linking with real threads libraries when thread safety
    constructor is enabled.

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>

Analysis:

    _XimRegisterIMInstantiateCallback() opens an XIM and closes it using
    the internal function pointers, but the internal close function does
    not free the pointer to the XIM (this would be done in XCloseIM()).

Report/patch:

    Date: Mon, 03 Oct 2022 18:47:32 +0800
    From: Po Lu <luangruo@yahoo.com>
    To: xorg-devel@lists.x.org
    Subject: Re: Yet another leak in Xlib

    For reference, here's how I'm calling XRegisterIMInstantiateCallback:

    XSetLocaleModifiers ("");
    XRegisterIMInstantiateCallback (compositor.display,
                                    XrmGetDatabase (compositor.display),
                                    (char *) compositor.resource_name,
                                    (char *) compositor.app_name,
                                    IMInstantiateCallback, NULL);

    and XMODIFIERS is:

        @im=ibus

Signed-off-by: Thomas E. Dickey <dickey@invisible-island.net>

Only really makes a difference if pthreads is not in libc.

Fixes: #162

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>
Reviewed-by: Adam Jackson <ajax@redhat.com>

Thomas E. Dickey authored 2 years ago and Thomas E. Dickey committed 2 years ago

Signed-off-by: Thomas E. Dickey <dickey@invisible-island.net>

Thomas E. Dickey authored 2 years ago and Thomas E. Dickey committed 2 years ago

Signed-off-by: Thomas E. Dickey <dickey@invisible-island.net>

Mike Fabian authored 2 years ago and Alan Coopersmith committed 2 years ago

https://gitlab.freedesktop.org/xorg/lib/libx11/-/blob/master/src/xlibi18n/lcCT.c#L58
has no escape sequence for KOI8-R.
That makes Xutf8TextListToTextProperty() sometimes fail when trying to
convert to COMPOUND_TEXT and when the KOI8-R charset is tried.

This is the cause for this bug:

https://github.com/ibus/ibus/issues/2422



Removing the KOI8-R charset entry fixes the problem.

This commit also fixes a few wrong texts in comments.

Signed-off-by: Mike FABIAN <mfabian@redhat.com>

Adam Sampson authored 3 years ago and Alan Coopersmith committed 2 years ago

The code here that made indexes greater than 3 refer to XKB symbol
groups had an off-by-one error, so it would always leave out the symbol
that should have been at index 4. Rewrite the code to fix this and
simplify the logic a bit.

Signed-off-by: Adam Sampson <ats@offog.org>

Signed-off-by: Thomas E. Dickey <dickey@invisible-island.net>

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>

Signed-off-by: David H. Gutteridge <david@gutteridge.ca>

Fixes: #160

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>

Fixes: #160

Signed-off-by: Alan Coopersmith <alan.coopersmith@oracle.com>

Jon Turney authored 2 years ago and Alan Coopersmith committed 2 years ago

/work/xorg/lib/libX11/src/XlibInt.c:1968: multiple definition of `_Xdebug_p'; .libs/globals.o:globals.c:(.bss+0xc): first defined here

Avoid redundant definition of _Xdebug_p in globals.c (which is under the
influence of _Xdebug being #defined to _Xdebug_p.

This appears to be an ancient hack to work around data exports resolving
to the address of the import stub, not the import. (See [1]).

(This is probably no longer needed or can be done in a better way, as
per the discussion under --enable-auto-import in the ld manpage.)

[1] https://cygwin.com/pipermail/cygwin-xfree/2001-May/004606.html



Signed-off-by: Jon Turney <jon.turney@dronecode.org.uk>

Benno Schulenberg authored 2 years ago and Alan Coopersmith committed 2 years ago

These sequences each produce two code points: the E-with-dot-above and
a combining macron.  The XIM input method is required for this to work.

(Also add a missing comment for a Unicode block.)

This fixes issue #54.

Requested-by: Arns Udovīčė

Signed-off-by: Benno Schulenberg <bensberg@telfort.nl>

Typing a compose sequence requires some care -- surely the user is able
to either keep holding the Shift key or not touch it at all while typing
the sequence.  Also, compose sequences are not an infinite resource AND
take up space and time -- defining redundant ones is a waste.

Signed-off-by: Benno Schulenberg <bensberg@telfort.nl>