compilers: convert `b_sanitize` to a free-form array option

In the preceding commit we have started to perform compiler checks for
the value of `b_sanitize`, which allows us to detect sanitizers that
aren't supported by the compiler toolchain. But we haven't yet loosened
the option itself to accept arbitrary values, so until now it's still
only possible to pass sanitizer combinations known by Meson, which is
quite restrictive.

Lift that restriction by adapting the `b_sanitize` option to become a
free-form array. Like this, users can pass whatever combination of
comma-separated sanitizers to Meson, which will then figure out whether
that combination is supported via the compiler checks. This lifts a
couple of restrictions and makes the supporting infrastructure way more
future proof.

A couple of notes regarding backwards compatibility:

  - All previous values of `b_sanitize` will remain valid as the syntax
    for free-form array values and valid combo choices is the same. We
    also treat 'none' specially so that we know to convert it into an
    empty array.

  - Even though the option has been converted into a free-form array,
    callers of `get_option('b_sanitize')` continue to get a string as
    value. We may eventually want to introduce a kwarg to alter this
    behaviour, but for now it is expected to be good enough for most use
    cases.

Fixes #8283
Fixes #7761
Fixes #5154
Fixes #1582

Co-authored-by: Dylan Baker <dylan@pnwbakers.com>
Signed-off-by: Patrick Steinhardt <ps@pks.im>

14 files changed, 119 insertions, 44 deletions

diff --git a/docs/markdown/Builtin-options.md b/docs/markdown/Builtin-options.md
index ffbab47..e1686f8 100644
--- a/docs/markdown/Builtin-options.md
+++ b/docs/markdown/Builtin-options.md
@@ -231,10 +231,20 @@ available on all platforms or with all compilers:
 | b_pie               | false                | true, false                                                   | Build position-independent executables (since 0.49.0)                          |
 | b_vscrt             | from_buildtype       | none, md, mdd, mt, mtd, from_buildtype, static_from_buildtype | VS runtime library to use (since 0.48.0) (static_from_buildtype since 0.56.0)  |
 
-The value of `b_sanitize` can be one of: `none`, `address`, `thread`,
-`undefined`, `memory`, `leak`, `address,undefined`, but note that some
-compilers might not support all of them. For example Visual Studio
-only supports the address sanitizer.
+The default and possible values of sanitizers changed in 1.8. Before 1.8 they
+were string values, and restricted to a specific subset of values: `none`,
+`address`, `thread`, `undefined`, `memory`, `leak`, or `address,undefined`. In
+1.8 it was changed to a free form array of sanitizers, which are checked by a
+compiler and linker check. For backwards compatibility reasons
+`get_option('b_sanitize')` continues to return a string with the array values
+separated by a comma. Furthermore:
+
+ - If the `b_sanitize` option is empty, the `'none'` string is returned.
+
+ - If it contains only the values `'address'` and `'undefined'`, they are
+   always returned as the `'address,undefined'` string, in this order.
+
+ - Otherwise, the array elements are returned in undefined order.
 
 \* < 0 means disable, == 0 means automatic selection, > 0 sets a specific number to use
 
diff --git a/docs/markdown/snippets/b_sanitizer_changes.md b/docs/markdown/snippets/b_sanitizer_changes.md
new file mode 100644
index 0000000..f726d70
--- /dev/null
+++ b/docs/markdown/snippets/b_sanitizer_changes.md
@@ -0,0 +1,17 @@
+## Changes to the b_sanitize option
+
+Before 1.8 the `b_sanitize` option was a combo option, which is an enumerated
+set of values. In 1.8 this was changed to a free-form array of options where
+available sanitizers are not hardcoded anymore but instead verified via a
+compiler check.
+
+This solves a number of longstanding issues such as:
+
+ - Sanitizers may be supported by a compiler, but not on a specific platform
+   (OpenBSD).
+ - New sanitizers are not recognized by Meson.
+ - Using sanitizers in previously-unsupported combinations.
+
+To not break backwards compatibility, calling `get_option('b_sanitize')`
+continues to return the configured value as a string, with a guarantee that
+`address,undefined` remains ordered.
diff --git a/mesonbuild/backend/vs2010backend.py b/mesonbuild/backend/vs2010backend.py
index 1015083..feef3a7 100644
--- a/mesonbuild/backend/vs2010backend.py
+++ b/mesonbuild/backend/vs2010backend.py
@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: Apache-2.0
 # Copyright 2014-2016 The Meson development team
+# Copyright © 2023-2024 Intel Corporation
 
 from __future__ import annotations
 import copy
@@ -272,7 +273,7 @@ class Vs2010Backend(backends.Backend):
         try:
             self.sanitize = self.environment.coredata.get_option(OptionKey('b_sanitize'))
         except KeyError:
-            self.sanitize = 'none'
+            self.sanitize = []
         sln_filename = os.path.join(self.environment.get_build_dir(), self.build.project_name + '.sln')
         projlist = self.generate_projects(vslite_ctx)
         self.gen_testproj()
diff --git a/mesonbuild/compilers/compilers.py b/mesonbuild/compilers/compilers.py
index 27bc44b..f0744f8 100644
--- a/mesonbuild/compilers/compilers.py
+++ b/mesonbuild/compilers/compilers.py
@@ -222,9 +222,7 @@ BASE_OPTIONS: T.Mapping[OptionKey, options.AnyOptionType] = {
         options.UserComboOption('b_lto_mode', 'Select between different LTO modes.', 'default', choices=['default', 'thin']),
         options.UserBooleanOption('b_thinlto_cache', 'Use LLVM ThinLTO caching for faster incremental builds', False),
         options.UserStringOption('b_thinlto_cache_dir', 'Directory to store ThinLTO cache objects', ''),
-        options.UserComboOption(
-            'b_sanitize', 'Code sanitizer to use', 'none',
-            choices=['none', 'address', 'thread', 'undefined', 'memory', 'leak', 'address,undefined']),
+        options.UserStringArrayOption('b_sanitize', 'Code sanitizer to use', []),
         options.UserBooleanOption('b_lundef', 'Use -Wl,--no-undefined when linking', True),
         options.UserBooleanOption('b_asneeded', 'Use -Wl,--as-needed when linking', True),
         options.UserComboOption(
@@ -307,7 +305,9 @@ def get_base_compile_args(target: 'BuildTarget', compiler: 'Compiler', env: 'Env
         pass
     try:
         sanitize = env.coredata.get_option_for_target(target, 'b_sanitize')
-        assert isinstance(sanitize, str)
+        assert isinstance(sanitize, list)
+        if sanitize == ['none']:
+            sanitize = []
         sanitize_args = compiler.sanitizer_compile_args(sanitize)
         # We consider that if there are no sanitizer arguments returned, then
         # the language doesn't support them.
@@ -376,7 +376,9 @@ def get_base_link_args(target: 'BuildTarget',
         pass
     try:
         sanitizer = env.coredata.get_option_for_target(target, 'b_sanitize')
-        assert isinstance(sanitizer, str)
+        assert isinstance(sanitizer, list)
+        if sanitizer == ['none']:
+            sanitizer = []
         sanitizer_args = linker.sanitizer_link_args(sanitizer)
         # We consider that if there are no sanitizer arguments returned, then
         # the language doesn't support them.
@@ -1044,10 +1046,10 @@ class Compiler(HoldableObject, metaclass=abc.ABCMeta):
                           thinlto_cache_dir: T.Optional[str] = None) -> T.List[str]:
         return self.linker.get_lto_args()
 
-    def sanitizer_compile_args(self, value: str) -> T.List[str]:
+    def sanitizer_compile_args(self, value: T.List[str]) -> T.List[str]:
         return []
 
-    def sanitizer_link_args(self, value: str) -> T.List[str]:
+    def sanitizer_link_args(self, value: T.List[str]) -> T.List[str]:
         return self.linker.sanitizer_args(value)
 
     def get_asneeded_args(self) -> T.List[str]:
diff --git a/mesonbuild/compilers/cuda.py b/mesonbuild/compilers/cuda.py
index 134cd4e..612643c 100644
--- a/mesonbuild/compilers/cuda.py
+++ b/mesonbuild/compilers/cuda.py
@@ -1,6 +1,6 @@
 # SPDX-License-Identifier: Apache-2.0
 # Copyright 2012-2017 The Meson development team
-# Copyright © 2024 Intel Corporation
+# Copyright © 2023-2024 Intel Corporation
 
 from __future__ import annotations
 
@@ -700,10 +700,10 @@ class CudaCompiler(Compiler):
         # return self._to_host_flags(self.host_compiler.get_optimization_args(optimization_level))
         return cuda_optimization_args[optimization_level]
 
-    def sanitizer_compile_args(self, value: str) -> T.List[str]:
+    def sanitizer_compile_args(self, value: T.List[str]) -> T.List[str]:
         return self._to_host_flags(self.host_compiler.sanitizer_compile_args(value))
 
-    def sanitizer_link_args(self, value: str) -> T.List[str]:
+    def sanitizer_link_args(self, value: T.List[str]) -> T.List[str]:
         return self._to_host_flags(self.host_compiler.sanitizer_link_args(value))
 
     def get_debug_args(self, is_debug: bool) -> T.List[str]:
diff --git a/mesonbuild/compilers/mixins/gnu.py b/mesonbuild/compilers/mixins/gnu.py
index 4dc3445..70fd9ee 100644
--- a/mesonbuild/compilers/mixins/gnu.py
+++ b/mesonbuild/compilers/mixins/gnu.py
@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: Apache-2.0
 # Copyright 2019-2022 The meson development team
+# Copyright © 2023 Intel Corporation
 
 from __future__ import annotations
 
@@ -495,11 +496,11 @@ class GnuLikeCompiler(Compiler, metaclass=abc.ABCMeta):
         # for their specific arguments
         return ['-flto']
 
-    def sanitizer_compile_args(self, value: str) -> T.List[str]:
-        if value == 'none':
-            return []
-        args = ['-fsanitize=' + value]
-        if 'address' in value:  # for -fsanitize=address,undefined
+    def sanitizer_compile_args(self, value: T.List[str]) -> T.List[str]:
+        if not value:
+            return value
+        args = ['-fsanitize=' + ','.join(value)]
+        if 'address' in value:
             args.append('-fno-omit-frame-pointer')
         return args
 
diff --git a/mesonbuild/compilers/mixins/islinker.py b/mesonbuild/compilers/mixins/islinker.py
index 44040a7..3f35619 100644
--- a/mesonbuild/compilers/mixins/islinker.py
+++ b/mesonbuild/compilers/mixins/islinker.py
@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: Apache-2.0
 # Copyright 2019 The Meson development team
+# Copyright © 2023 Intel Corporation
 
 from __future__ import annotations
 
@@ -37,7 +38,7 @@ class BasicLinkerIsCompilerMixin(Compiler):
     functionality itself.
     """
 
-    def sanitizer_link_args(self, value: str) -> T.List[str]:
+    def sanitizer_link_args(self, value: T.List[str]) -> T.List[str]:
         return []
 
     def get_lto_link_args(self, *, threads: int = 0, mode: str = 'default',
diff --git a/mesonbuild/compilers/mixins/visualstudio.py b/mesonbuild/compilers/mixins/visualstudio.py
index 30127ec..275e7ab 100644
--- a/mesonbuild/compilers/mixins/visualstudio.py
+++ b/mesonbuild/compilers/mixins/visualstudio.py
@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: Apache-2.0
 # Copyright 2019 The meson development team
+# Copyright © 2023 Intel Corporation
 
 from __future__ import annotations
 
@@ -166,12 +167,10 @@ class VisualStudioLikeCompiler(Compiler, metaclass=abc.ABCMeta):
     def get_no_optimization_args(self) -> T.List[str]:
         return ['/Od', '/Oi-']
 
-    def sanitizer_compile_args(self, value: str) -> T.List[str]:
-        if value == 'none':
-            return []
-        if value != 'address':
-            raise mesonlib.MesonException('VS only supports address sanitizer at the moment.')
-        return ['/fsanitize=address']
+    def sanitizer_compile_args(self, value: T.List[str]) -> T.List[str]:
+        if not value:
+            return value
+        return [f'/fsanitize={",".join(value)}']
 
     def get_output_args(self, outputname: str) -> T.List[str]:
         if self.mode == 'PREPROCESSOR':
diff --git a/mesonbuild/interpreter/interpreter.py b/mesonbuild/interpreter/interpreter.py
index ff93ac6..42e6243 100644
--- a/mesonbuild/interpreter/interpreter.py
+++ b/mesonbuild/interpreter/interpreter.py
@@ -1068,14 +1068,14 @@ class Interpreter(InterpreterBase, HoldableObject):
 
     @typed_pos_args('get_option', str)
     @noKwargs
-    def func_get_option(self, nodes: mparser.BaseNode, args: T.Tuple[str],
-                        kwargs: 'TYPE_kwargs') -> T.Union[options.UserOption, 'TYPE_var']:
+    def func_get_option(self, node: mparser.BaseNode, args: T.Tuple[str],
+                        kwargs: kwtypes.FuncGetOption) -> T.Union[options.UserOption, 'TYPE_var']:
         optname = args[0]
+
         if ':' in optname:
             raise InterpreterException('Having a colon in option name is forbidden, '
                                        'projects are not allowed to directly access '
                                        'options of other subprojects.')
-
         if optname_regex.search(optname.split('.', maxsplit=1)[-1]) is not None:
             raise InterpreterException(f'Invalid option name {optname!r}')
 
@@ -1096,6 +1096,15 @@ class Interpreter(InterpreterBase, HoldableObject):
             ocopy.name = optname
             ocopy.value = value
             return ocopy
+        elif optname == 'b_sanitize':
+            assert isinstance(value_object, options.UserStringArrayOption)
+            # To ensure backwards compatibility this always returns a string.
+            # We may eventually want to introduce a new "format" kwarg that
+            # allows the user to modify this behaviour, but for now this is
+            # likely good enough for most usecases.
+            if not value:
+                return 'none'
+            return ','.join(sorted(value))
         elif isinstance(value_object, options.UserOption):
             if isinstance(value_object.value, str):
                 return P_OBJ.OptionString(value, f'{{{optname}}}')
@@ -3090,7 +3099,7 @@ class Interpreter(InterpreterBase, HoldableObject):
         if OptionKey('b_sanitize') not in self.coredata.optstore:
             return
         if (self.coredata.optstore.get_value('b_lundef') and
-                self.coredata.optstore.get_value('b_sanitize') != 'none'):
+                self.coredata.optstore.get_value('b_sanitize')):
             value = self.coredata.optstore.get_value('b_sanitize')
             mlog.warning(textwrap.dedent(f'''\
                     Trying to use {value} sanitizer on Clang with b_lundef.
diff --git a/mesonbuild/linkers/linkers.py b/mesonbuild/linkers/linkers.py
index 0dc2c0b..b114d49 100644
--- a/mesonbuild/linkers/linkers.py
+++ b/mesonbuild/linkers/linkers.py
@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: Apache-2.0
 # Copyright 2012-2022 The Meson development team
+# Copyright © 2023 Intel Corporation
 
 from __future__ import annotations
 
@@ -223,7 +224,7 @@ class DynamicLinker(metaclass=abc.ABCMeta):
     def get_thinlto_cache_args(self, path: str) -> T.List[str]:
         return []
 
-    def sanitizer_args(self, value: str) -> T.List[str]:
+    def sanitizer_args(self, value: T.List[str]) -> T.List[str]:
         return []
 
     def get_asneeded_args(self) -> T.List[str]:
@@ -599,6 +600,9 @@ class PosixDynamicLinkerMixin(DynamicLinkerBase):
     def get_search_args(self, dirname: str) -> T.List[str]:
         return ['-L' + dirname]
 
+    def sanitizer_args(self, value: T.List[str]) -> T.List[str]:
+        return []
+
 
 class GnuLikeDynamicLinkerMixin(DynamicLinkerBase):
 
@@ -654,10 +658,10 @@ class GnuLikeDynamicLinkerMixin(DynamicLinkerBase):
     def get_lto_args(self) -> T.List[str]:
         return ['-flto']
 
-    def sanitizer_args(self, value: str) -> T.List[str]:
-        if value == 'none':
-            return []
-        return ['-fsanitize=' + value]
+    def sanitizer_args(self, value: T.List[str]) -> T.List[str]:
+        if not value:
+            return value
+        return [f'-fsanitize={",".join(value)}']
 
     def get_coverage_args(self) -> T.List[str]:
         return ['--coverage']
@@ -811,10 +815,10 @@ class AppleDynamicLinker(PosixDynamicLinkerMixin, DynamicLinker):
     def get_coverage_args(self) -> T.List[str]:
         return ['--coverage']
 
-    def sanitizer_args(self, value: str) -> T.List[str]:
-        if value == 'none':
-            return []
-        return ['-fsanitize=' + value]
+    def sanitizer_args(self, value: T.List[str]) -> T.List[str]:
+        if not value:
+            return value
+        return [f'-fsanitize={",".join(value)}']
 
     def no_undefined_args(self) -> T.List[str]:
         # We used to emit -undefined,error, but starting with Xcode 15 /
diff --git a/mesonbuild/modules/gnome.py b/mesonbuild/modules/gnome.py
index dffc615..e3b1f3d 100644
--- a/mesonbuild/modules/gnome.py
+++ b/mesonbuild/modules/gnome.py
@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: Apache-2.0
 # Copyright 2015-2016 The Meson development team
+# Copyright © 2023-2024 Intel Corporation
 
 '''This module provides helper functions for Gnome/GLib related
 functionality such as gobject-introspection, gresources and gtk-doc'''
@@ -912,9 +913,8 @@ class GnomeModule(ExtensionModule):
                 cflags += state.project_args[lang]
             if OptionKey('b_sanitize') in compiler.base_options:
                 sanitize = state.environment.coredata.optstore.get_value('b_sanitize')
-                assert isinstance(sanitize, str)
+                assert isinstance(sanitize, list)
                 cflags += compiler.sanitizer_compile_args(sanitize)
-                sanitize = sanitize.split(',')
                 # These must be first in ldflags
                 if 'address' in sanitize:
                     internal_ldflags += ['-lasan']
diff --git a/test cases/unit/125 sanitizers/meson.build b/test cases/unit/125 sanitizers/meson.build
new file mode 100644
index 0000000..b42fb35
--- /dev/null
+++ b/test cases/unit/125 sanitizers/meson.build
@@ -0,0 +1,8 @@
+# SPDX-License-Identifier: Apache-2.0
+# Copyright © 2023-2024 Intel Corporation
+
+project('sanitizer', 'c', meson_version : '>= 1.8')
+
+summary({
+  'value': get_option('b_sanitize'),
+}, section: 'summary')
diff --git a/unittests/allplatformstests.py b/unittests/allplatformstests.py
index 7f4e1e7..84d5245 100644
--- a/unittests/allplatformstests.py
+++ b/unittests/allplatformstests.py
@@ -1,5 +1,6 @@
 # SPDX-License-Identifier: Apache-2.0
 # Copyright 2016-2021 The Meson development team
+# Copyright © 2023-2024 Intel Corporation
 
 import subprocess
 import re
@@ -2770,7 +2771,7 @@ class AllPlatformTests(BasePlatformTests):
             obj = mesonbuild.coredata.load(self.builddir)
             self.assertEqual(obj.optstore.get_value('bindir'), 'bar')
             self.assertEqual(obj.optstore.get_value('buildtype'), 'release')
-            self.assertEqual(obj.optstore.get_value('b_sanitize'), 'thread')
+            self.assertEqual(obj.optstore.get_value('b_sanitize'), ['thread'])
             self.assertEqual(obj.optstore.get_value(OptionKey('c_args')), ['-Dbar'])
             self.setconf(['--bindir=bar', '--bindir=foo',
                           '-Dbuildtype=release', '-Dbuildtype=plain',
@@ -2779,7 +2780,7 @@ class AllPlatformTests(BasePlatformTests):
             obj = mesonbuild.coredata.load(self.builddir)
             self.assertEqual(obj.optstore.get_value('bindir'), 'foo')
             self.assertEqual(obj.optstore.get_value('buildtype'), 'plain')
-            self.assertEqual(obj.optstore.get_value('b_sanitize'), 'address')
+            self.assertEqual(obj.optstore.get_value('b_sanitize'), ['address'])
             self.assertEqual(obj.optstore.get_value(OptionKey('c_args')), ['-Dfoo'])
             self.wipe()
         except KeyError:
diff --git a/unittests/linuxliketests.py b/unittests/linuxliketests.py
index 08eb4b9..01862b6 100644
--- a/unittests/linuxliketests.py
+++ b/unittests/linuxliketests.py
@@ -1930,3 +1930,25 @@ class LinuxlikeTests(BasePlatformTests):
             self.check_has_flag(compdb, mainsrc, '-O3')
             self.check_has_flag(compdb, sub1src, '-O2')
             self.check_has_flag(compdb, sub2src, '-O2')
+
+    def test_sanitizers(self):
+        testdir = os.path.join(self.unit_test_dir, '125 sanitizers')
+
+        with self.subTest('no b_sanitize value'):
+            try:
+                out = self.init(testdir)
+                self.assertRegex(out, 'value *: *none')
+            finally:
+                self.wipe()
+
+        for value, expected in { '': 'none',
+                                 'none': 'none',
+                                 'address': 'address',
+                                 'undefined,address': 'address,undefined',
+                                 'address,undefined': 'address,undefined' }.items():
+            with self.subTest('b_sanitize=' + value):
+                try:
+                    out = self.init(testdir, extra_args=['-Db_sanitize=' + value])
+                    self.assertRegex(out, 'value *: *' + expected)
+                finally:
+                    self.wipe()