Removed an outdated performance note comment before the cached match state variables and a redundant comment in match_nick(); behavior remains unchanged.
Changed the flow to prefer xdg-open first (with sanitized env) and only then fall back to gtk_show_uri, instead of trying GTK first under a potentially broken runtime environment.
Added explicit final warning when both mechanisms fail, and ensured temporary allocations (including xdg_open_path) are released cleanly.
Kept the existing behavior of trying gtk_show_uri first and only using xdg-open as fallback; this change only hardens the fallback path and preserves existing flow/logging.
Added an xdg-open fallback path when gtk_show_uri fails, which improves behavior in AppImage/sandbox-like environments where GTK/GIO handler resolution can fail.
Added warning logs for both the primary gtk_show_uri failure and fallback failure to make future debugging much easier.
I also exposed global Lua aliases xchat and hexchat that both point to the zoitechat module table, so existing scripts calling xchat.* or hexchat.* work unchanged.
Added hexchat.py to the Visual Studio Python plugin project item list so it is tracked consistently in the Windows project metadata.
Added hexchat.py to the corresponding .vcxproj.filters entries under Source Files for IDE consistency.
Expanded ignored codepoints during highlight token comparison to include zero-width joiner (U+200D) in addition to variation selectors (U+FE0E/U+FE0F), which helps emoji entered from different input methods still match configured “extra highlight words.”
This logic is used by the alert/highlight matching path (alert_match_word) that compares configured extra words against extracted message tokens.
Added a Python hexchat.py compatibility module and ensured it is installed alongside zoitechat.py/xchat.py, enabling import hexchat scripts to run.
Updated the Perl API layer to support HexChat naming by aliasing HexChat:: to ZoiteChat::, adding EAT_HEXCHAT to exported constants, and adding an explicit Perl shim module HexChat.pm.
Extended Perl XS constant registration to export EAT_HEXCHAT as a compatibility alias, and wired HexChat.pm into the embedded Perl module generation list so it ships with the plugin.
Improved token scanning to use UTF-8-aware character classification (gunichar, g_unichar_isdigit, g_unichar_isalpha) instead of byte-only checks, avoiding split/mis-detection on multibyte characters.
Added a direct find_program() lookup for glib-genmarshal (with glib-genmarshal-2.0 fallback), so configuration no longer depends on the broken glib-2.0 pkg-config tool variable (glib_genmarshal) reported in your error.
Added a new GitHub Actions workflow .github/workflows/manjaro-package-build.yml that:
runs in a manjarolinux/base container,
installs Manjaro packaging/build tooling,
builds via makepkg as a non-root user,
generates .SRCINFO,
runs namcap checks on both PKGBUILD and the built package,
uploads package artifacts plus PKGBUILD/.SRCINFO.
narrowing the GDK include guard to #if defined(HAVE_GTK3), and
removing legacy HAVE_GTK2 / HAVE_GTK toolkit label branches from sysinfo_detect_toolkit(),
narrowing GDK display probing guard in sysinfo_detect_display_backend() to GTK3-only.
Updated Windows and macOS sysinfo backends to remove HAVE_GTK2 / HAVE_GTK fallback labels so sysinfo_detect_toolkit() now returns only "GTK3" (or NULL).
Updated API docs for sysinfo_backend_get_ui() to remove the outdated "GTK2 / X11" example so comments match runtime behavior.
Removed the deprecated Meson option option('gtk3', ... deprecated: true) from meson_options.txt, leaving gtk-frontend/text-frontend options intact.
Removed -Dgtk3=true from Debian’s Meson configure flags in debian/rules (override_dh_auto_configure).
Removed remaining active -Dgtk3=true packaging consumers in Flatpak config:
dropped it from config-opts
dropped -Dgtk3=true from cflags.
Removed -Dgtk3=true from Solus packaging Meson configure args.
Committed changes on the current branch (e018bfd) and created a PR record via the make_pr tool.
