From 588643f6a8abbd10a0c704b4d60f26db7d28d898 Mon Sep 17 00:00:00 2001
From: David Seifert <soap@gentoo.org>
Date: Wed, 22 May 2024 15:55:55 +0200
Subject: [PATCH] cuda: fix `cuda.find_library()` hardcoded to yield true

* Previously, cuda would just plainly prepend `-l` to the libname.
* By relying on the host compiler to find libraries, we now get
  more subtle failures, such as CUDA modules not being found
  anymore.
* We need to simplify these CUDA modules when nvcc is used for
  linking, since this may have side-effects from the cuda toolchain.

Closes: #13240
---
 mesonbuild/compilers/compilers.py |  2 +-
 mesonbuild/compilers/cuda.py      |  2 +-
 mesonbuild/dependencies/cuda.py   | 43 ++++++++++++++++++++-----------
 3 files changed, 30 insertions(+), 17 deletions(-)

diff --git a/mesonbuild/compilers/compilers.py b/mesonbuild/compilers/compilers.py
index 44b998a67cda..673b63f9d581 100644
--- a/mesonbuild/compilers/compilers.py
+++ b/mesonbuild/compilers/compilers.py
@@ -1046,7 +1046,7 @@ def get_dependency_compile_args(self, dep: 'Dependency') -> T.List[str]:
         return dep.get_compile_args()
 
     def get_dependency_link_args(self, dep: 'Dependency') -> T.List[str]:
-        return dep.get_link_args()
+        return dep.get_link_args(self.get_language())
 
     @classmethod
     def use_linker_args(cls, linker: str, version: str) -> T.List[str]:
diff --git a/mesonbuild/compilers/cuda.py b/mesonbuild/compilers/cuda.py
index 3761019b9945..7231a267c02b 100644
--- a/mesonbuild/compilers/cuda.py
+++ b/mesonbuild/compilers/cuda.py
@@ -768,7 +768,7 @@ def get_std_exe_link_args(self) -> T.List[str]:
 
     def find_library(self, libname: str, env: 'Environment', extra_dirs: T.List[str],
                      libtype: LibType = LibType.PREFER_SHARED, lib_prefix_warning: bool = True) -> T.Optional[T.List[str]]:
-        return ['-l' + libname] # FIXME
+        return self.host_compiler.find_library(libname, env, extra_dirs, libtype, lib_prefix_warning)
 
     def get_crt_compile_args(self, crt_val: str, buildtype: str) -> T.List[str]:
         return self._to_host_flags(self.host_compiler.get_crt_compile_args(crt_val, buildtype))
diff --git a/mesonbuild/dependencies/cuda.py b/mesonbuild/dependencies/cuda.py
index a38e325b8741..82bf5ad82b94 100644
--- a/mesonbuild/dependencies/cuda.py
+++ b/mesonbuild/dependencies/cuda.py
@@ -41,11 +41,6 @@ def __init__(self, environment: 'Environment', kwargs: T.Dict[str, T.Any]) -> No
             req_modules = ['cudart']
             if kwargs.get('static', True):
                 req_modules = ['cudart_static']
-                machine = self.env.machines[self.for_machine]
-                if machine.is_linux():
-                    # extracted by running
-                    #   nvcc -v foo.o
-                    req_modules += ['rt', 'pthread', 'dl']
             self.requested_modules = req_modules + self.requested_modules
 
         (self.cuda_path, self.version, self.is_found) = self._detect_cuda_path_and_version()
@@ -61,12 +56,9 @@ def __init__(self, environment: 'Environment', kwargs: T.Dict[str, T.Any]) -> No
             self.incdir = os.path.join(self.cuda_path, 'include')
             self.compile_args += [f'-I{self.incdir}']
 
-        if self.language != 'cuda':
-            arch_libdir = self._detect_arch_libdir()
-            self.libdir = os.path.join(self.cuda_path, arch_libdir)
-            mlog.debug('CUDA library directory is', mlog.bold(self.libdir))
-        else:
-            self.libdir = None
+        arch_libdir = self._detect_arch_libdir()
+        self.libdir = os.path.join(self.cuda_path, arch_libdir)
+        mlog.debug('CUDA library directory is', mlog.bold(self.libdir))
 
         self.is_found = self._find_requested_libraries()
 
@@ -244,7 +236,14 @@ def _find_requested_libraries(self) -> bool:
         all_found = True
 
         for module in self.requested_modules:
-            args = self.clib_compiler.find_library(module, self.env, [self.libdir] if self.libdir else [])
+            args = self.clib_compiler.find_library(module, self.env, [self.libdir])
+            if module == 'cudart_static' and self.language != 'cuda':
+                machine = self.env.machines[self.for_machine]
+                if machine.is_linux():
+                    # extracted by running
+                    #   nvcc -v foo.o
+                    args += ['-lrt', '-lpthread', '-ldl']
+
             if args is None:
                 self._report_dependency_error(f'Couldn\'t find requested CUDA module \'{module}\'')
                 all_found = False
@@ -286,10 +285,24 @@ def get_requested(self, kwargs: T.Dict[str, T.Any]) -> T.List[str]:
 
     def get_link_args(self, language: T.Optional[str] = None, raw: bool = False) -> T.List[str]:
         args: T.List[str] = []
-        if self.libdir:
-            args += self.clib_compiler.get_linker_search_args(self.libdir)
         for lib in self.requested_modules:
-            args += self.lib_modules[lib]
+            link_args = self.lib_modules[lib]
+            # Turn canonical arguments like
+            #   /opt/cuda/lib64/libcublas.so
+            # back into
+            #   -lcublas
+            # since this is how CUDA modules were passed to nvcc since time immemorial
+            if language == 'cuda':
+                if lib in frozenset(['cudart', 'cudart_static']):
+                    # nvcc always links these unconditionally
+                    mlog.debug(f'Not adding \'{lib}\' to dependency, since nvcc will link it implicitly')
+                    link_args = []
+                elif link_args and link_args[0].startswith(self.libdir):
+                    # module included with CUDA, nvcc knows how to find these itself
+                    mlog.debug(f'CUDA module \'{lib}\' found in CUDA libdir')
+                    link_args = ['-l' + lib]
+            args += link_args
+
         return args
 
 packages['cuda'] = CudaDependency