ci: Set some ubsan options

Hopefully, this makes the build survive.
Drop -Werror=array-bounds for now
2024-03-15 00:38:12 -04:00 · 2024-03-15 00:14:17 -04:00 · 2024-03-14 23:33:29 -04:00 · 2024-03-14 22:32:43 -04:00 · 2024-03-14 22:32:24 -04:00 · 2024-03-14 15:30:51 -04:00
7 changed files with 7 additions and 8 deletions
--- a/.gitlab-ci.yml
+++ b/.gitlab-ci.yml
@@ -26,7 +26,7 @@ variables:
  BACKEND_FLAGS: "-Dx11-backend=true -Dwayland-backend=true -Dbroadway-backend=true"
  FEATURE_FLAGS: "-Dvulkan=enabled -Dcloudproviders=enabled -Dbuild-testsuite=true -Dintrospection=enabled"
  MESON_TEST_TIMEOUT_MULTIPLIER: 3
-  FEDORA_IMAGE: "registry.gitlab.gnome.org/gnome/gtk/fedora:v49"
+  FEDORA_IMAGE: "registry.gitlab.gnome.org/gnome/gtk/fedora:v50"

 workflow:
  rules:
@@ -422,7 +422,7 @@ asan-build:
    - export PATH="$HOME/.local/bin:$PATH"
    - CC=clang meson setup
            --buildtype=debugoptimized
-            -Db_sanitize=address
+            -Db_sanitize=address,undefined
            -Db_lundef=false
            -Dbuild-demos=false
            -Dbuild-tests=false
--- a/.gitlab-ci/.fedora.Dockerfile.swp
+++ b/.gitlab-ci/.fedora.Dockerfile.swp
--- a/.gitlab-ci/fedora.Dockerfile
+++ b/.gitlab-ci/fedora.Dockerfile
@@ -2,8 +2,6 @@ FROM fedora:39

 RUN dnf -y install \
    adwaita-icon-theme \
-    atk-devel \
-    at-spi2-atk-devel \
    avahi-gobject-devel \
    cairo-devel \
    cairo-gobject-devel \
@@ -18,7 +16,6 @@ RUN dnf -y install \
    dejavu-sans-mono-fonts \
    desktop-file-utils \
    diffutils \
-    docbook-style-xsl \
    elfutils-libelf-devel \
    expat-devel \
    fribidi-devel \
--- a/.gitlab-ci/run-tests.sh
+++ b/.gitlab-ci/run-tests.sh
@@ -11,7 +11,7 @@ multiplier=${MESON_TEST_TIMEOUT_MULTIPLIER:-1}

 # Ignore memory leaks lower in dependencies
 export LSAN_OPTIONS=suppressions=$srcdir/lsan.supp:print_suppressions=0:detect_leaks=0:allocator_may_return_null=1
-export G_SLICE=always-malloc
+export UBSAN_OPTIONS=silence_unsigned_overflow=1

 case "${setup}" in
  x11*)
--- a/gdk/gdkglcontext.c
+++ b/gdk/gdkglcontext.c
@@ -290,7 +290,7 @@ gdk_gl_context_create_egl_context (GdkGLContext *context,
  GdkDisplay *display = gdk_gl_context_get_display (context);
  EGLDisplay egl_display = gdk_display_get_egl_display (display);
  GdkGLContext *share = gdk_display_get_gl_context (display);
-  GdkGLContextPrivate *share_priv = gdk_gl_context_get_instance_private (share);
+  GdkGLContextPrivate *share_priv = share ? gdk_gl_context_get_instance_private (share) : NULL;
  EGLConfig egl_config;
  EGLContext ctx;
  EGLint context_attribs[N_EGL_ATTRS], i = 0, flags = 0;
--- a/meson.build
+++ b/meson.build
@@ -307,7 +307,6 @@ elif cc.get_id() == 'gcc' or cc.get_id() == 'clang'

  extra_warnings = [
    'address',
-    'array-bounds',
    'empty-body',
    'enum-int-mismatch',
    'implicit',
--- a/testsuite/tools/validate
+++ b/testsuite/tools/validate
@@ -4,6 +4,9 @@ GTK_BUILDER_TOOL=${GTK_BUILDER_TOOL:-gtk4-builder-tool}
 TEST_DATA_DIR=${G_TEST_SRCDIR:-.}/validate-data
 TEST_RESULT_DIR=${TEST_RESULT_DIR:-/tmp}/validate

+# https://gitlab.freedesktop.org/mesa/mesa/-/issues/10293
+export LIBGL_ALWAYS_SOFTWARE=true
+
 mkdir -p "$TEST_RESULT_DIR"

 shopt -s nullglob
Author	SHA1	Message	Date
Matthias Clasen	79b30bdf0a	ci: Set some ubsan options Hopefully, this makes the build survive.	2024-03-15 00:38:12 -04:00
Matthias Clasen	b832b3fd35	Drop -Werror=array-bounds for now Until somebody fixes the code in gtkcssselector.c in a way that makes the compiler happy, don't pass unhappiness-inducing flags.	2024-03-15 00:14:17 -04:00
Matthias Clasen	c0391ce193	glcontext: Don't call get_instance_private (NULL) ubsan does not like this. See `a9175e0c03` for a detailed explanation.	2024-03-14 23:33:29 -04:00
Simon McVittie	05d36b2ba4	testsuite: Force Mesa to use software GL when asserting about stderr In an autobuilder environment, there will typically be no hardware GPU available, so Mesa will fall back from hardware to Zink to software rendering. Unfortunately, Zink logs to stderr during loading if no hardware GPUs are available. This particular test asserts that stderr has desired contents, which means Zink's extra output causes the test to fail. We can bypass this by disabling use of Zink. Resolves: https://gitlab.gnome.org/GNOME/gtk/-/issues/6478 Signed-off-by: Simon McVittie <smcv@debian.org>	2024-03-14 22:32:43 -04:00
Matthias Clasen	a904292bfb	Drop an unintenionally added file	2024-03-14 22:32:24 -04:00
Matthias Clasen	65cf59408b	ci: Use v50 of the Fedora image This has dropped some unnecessary packages.	2024-03-14 15:30:51 -04:00
Matthias Clasen	b269a4712a	ci: Use ubsan as well We have libubsan in the ci image, so we may as well enable both asan and ubsan.	2024-03-14 11:39:51 -04:00
Matthias Clasen	ebd74e112d	ci: Drop some unnecessary packages We don't use at, at-spi or the docbook stylesheets.	2024-03-14 11:37:55 -04:00