[BUG] Fix clang-tidy errors

[teju85]

Describe the bug
In the past we had to disable parsing of .cu files in run-clang-tidy.py due to difficulties in trying to get them working properly. So, we never tidied our cuda source files!

Steps/Code to reproduce bug
Today, if we path this script with the following change:

diff --git a/cpp/scripts/run-clang-tidy.py b/cpp/scripts/run-clang-tidy.py
index 23260d2..fdb0abc 100644
--- a/cpp/scripts/run-clang-tidy.py
+++ b/cpp/scripts/run-clang-tidy.py
@@ -36,7 +36,7 @@ def parse_args():
                            help="Path to cmake-generated compilation database")
     argparser.add_argument("-exe", type=str, default="clang-tidy",
                            help="Path to clang-tidy exe")
-    argparser.add_argument("-ignore", type=str, default="[.]cu$",
+    argparser.add_argument("-ignore", type=str, default=None,
                            help="Regex used to ignore files from checking")
     argparser.add_argument("-select", type=str, default=None,
                            help="Regex used to select files for checking")
@@ -123,6 +123,10 @@ def get_tidy_args(cmd, exe):
         # replace nvcc's "-gencode ..." with clang's "--cuda-gpu-arch ..."
         archs = get_gpu_archs(command)
         command.extend(archs)
+        # clang-tooling 8.0.1 doesn't support linking with greater sm_70 archs
+        # so, we need to disable libdevice linkage, until we are in a position
+        # to upgrade clang-tooling
+        command.extend(["-nocudalib", "--no-cuda-version-check"])
         while True:
             loc = remove_item_plus_one(command, "-gencode")
             if loc < 0:

and then run the script, we get a ton of errors which we never caught before (and thus never fixed). Full list of all the errors in here: run.log

Expected behavior
tidy check should pass on all source files, irrespective of device or host code.

[github-actions]

This issue has been marked rotten due to no recent activity in the past 90d. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed.

[github-actions]

This issue has been marked stale due to no recent activity in the past 30d. Please close this issue if no further response or action is needed. Otherwise, please respond with a comment indicating any updates or changes to the original issue and/or confirm this issue still needs to be addressed. This issue will be marked rotten if there is no activity in the next 60d.

[MatthiasKohl]

I'd like to come back to this issue and fix clang compilation in raft.
There are several things that I can see after attempting to compile all raft test codes with clang:

1. A few minor things due to the CUDA/C __host__ / __device__ overloads which nvcc and clang++ treat differently (mainly functions such as min / fmin etc.)
2. The fact that raft has a dependency on cub which is too old to be compiled by clang: we should update this to >= 1.14
3. A few issues with static const vs static constexpr : should be easy to handle since raft is C++ >= 11
4. There seem to be some issues with template parameters in KNN which I still have to figure out