Add conversions from G_GINT64_CONSTANT to INT64_C and G_GUINT64_CONSTANT
to UINT64_C to convert-glib-types. Convert instances in epan/ftypes and
wsutil.
Patch lemon to stop parsing options upon encountering "--".
This prevents an issue in the unlikely situation where the
source directory or a source filename has a '=' in it, or
starts with a "-".
Redo 3716933046 but only
use the -- option in our lemon, not when using system lemon.
Patch lemon to stop parsing options upon encountering "--".
This prevents an issue in the unlikely situation where the
source directory or a source filename has a '=' in it, or
starts with a "-".
This has been reported to sqlite upstream.
Instead of adding a TFS(&tfs_true_false) to every boolean
field, make it the default if "strings" is NULL.
This seems to match the already existing documentation:
If the Boolean field is to be displayed as "False" or "True", the
'strings' field would be set to NULL.
Libraries not built with CMake have as their install_path (meaning "the
install path recorded in the dylib file", not "the path to which the
file is installed") the full path of the library where it's installed.
CMake, at least in newer versions, defaults to @rpath/{library file name}.
This causes binaries built in the Wireshark binary process, and linked
with those libraries, not to be executable from the build directory, at
least on macOS Sonoma when built with Xcode 15. (See comment in the
script for details.)
Define DO_CMAKE to run cmake with -D options to force libraries to have
their install_path be the full path to which the library is installed.
Before checking whether we need to install a .pc file for libffi, so
that GLib's configuration process finds the libffi that Apple has
shipped with at least some macOS releases (library) and Xcode releases
(headers), check whether we have one that's out-of-date due to pointing
to an SDK that was installed the last time we ran the script but that's
not installed now due to a newer Xcode release having been installed in
the interim.
(Yes, this can happen, and, yes, this did happen, causing mysterious
build failures due to code in Glib trying to include <ffi.h> rather than
<ffi/ffi.h>, as they expected libffi's .pc file to point to an ffi
include directory that exists.)
Sonoma replaced the GNU libiconv with the FreeBSD iconv code; it
conforms to POSIX, but does not have the same behavior as GNU libiconv
in a case where POSIX doesn't specify the behavior.
The configure script for GNU gettext tests for GNU behavior and doesn't
use iconv if it sees non-GNU behavior; unfortunately, that causes the
build of gettext to break. (See the comment for the details.)
Work around it the same way Homebrew does.
Attempt to fix the following CI job failure:
$ mingw64-cmake -G Ninja -DENABLE_CCACHE=Yes -DFETCH_lua=Yes ..
/usr/bin/mingw64-cmake: line 91: fg: no job control
[skip ci]
* None of the OUI tables are supposed to be written to, constify them.
* Use proper types in the bsearch parameters to avoid confusion.
* Move masking outside the bsearch function as tiny optimization.
* Document the MA-L/M/S macros.