[libc++] Improve the format ignorelist generation.

[Mordante]

Summary

Several improvements

• Only add files that we actually want to format.
• Sort according to a fixed locale.

Some drive-by fixes

• Rename a text file, this avoids a filter exception.
• Adds a some missing source files extensions.
• Removes unused extensions hh, hxx, cc, and cxx from clang-format.

[Mordante]

Remove testing part and minor improvements.

[philnik]

LGTM % comment. Thanks for improving the script.

libcxx/utils/generate_ignore_format.sh

16

AFAIK we don't have most of these file endings. IMO we shouldn't list them here (and remove them from check-format).

[Mordante]

Thanks for the review.

libcxx/utils/generate_ignore_format.sh

16

*.c is unused, but I'm want to keep it.
*.hh, *.hxx, *.cc, and *.cxx are unused and these I'll remove.

[philnik]

libcxx/utils/generate_ignore_format.sh

16

Sounds good.

libcxx/utils/ci/run-buildbot

check-format)
     echo "+++ Checking formatting"
     # We need to set --extensions so that clang-format checks extensionless files.
     mkdir -p ${BUILD_DIR}
     if [ -z "${GIT_CLANG_FORMAT}" ]; then
        GIT_CLANG_FORMAT=git-clang-format
     fi
     ${GIT_CLANG_FORMAT} \
         --diff \
-        --extensions ',h,hh,hpp,hxx,c,cc,cxx,cpp' HEAD~1 \
+        --extensions ',h,hpp,c,cpp,inc,ipp' HEAD~1 \
         -- $(find libcxx/{benchmarks,include,src}/ -type f | grep -vf libcxx/utils/data/ignore_format.txt) \
         | tee ${BUILD_DIR}/clang-format.patch
     # Check if the diff is empty, fail otherwise.
     ! grep -q '^--- a' ${BUILD_DIR}/clang-format.patch
 ;;
 check-generated-output)
     # `! foo` doesn't work properly with `set -e`, use `! foo || false` instead.
     # https://stackoverflow.com/questions/57681955/set-e-does-not-respect-logical-not

libcxx/utils/data/ignore_format.txt

-libcxx/benchmarks/CMakeLists.txt
 libcxx/benchmarks/CartesianBenchmarks.h
 libcxx/benchmarks/ContainerBenchmarks.h
 libcxx/benchmarks/GenerateInput.h
 libcxx/benchmarks/Utilities.h
 libcxx/benchmarks/VariantBenchmarks.h
 libcxx/benchmarks/algorithms/common.h
 libcxx/benchmarks/algorithms/lower_bound.bench.cpp
 libcxx/benchmarks/algorithms/make_heap.bench.cpp

 libcxx/benchmarks/allocation.bench.cpp
 libcxx/benchmarks/deque.bench.cpp
 libcxx/benchmarks/filesystem.bench.cpp
 libcxx/benchmarks/formatter_float.bench.cpp
 libcxx/benchmarks/formatter_int.bench.cpp
 libcxx/benchmarks/format_to.bench.cpp
 libcxx/benchmarks/format_to_n.bench.cpp
 libcxx/benchmarks/function.bench.cpp
-libcxx/benchmarks/lit.cfg.py
-libcxx/benchmarks/lit.site.cfg.py.in
 libcxx/benchmarks/map.bench.cpp
 libcxx/benchmarks/ordered_set.bench.cpp
 libcxx/benchmarks/random.bench.cpp
 libcxx/benchmarks/string.bench.cpp
 libcxx/benchmarks/stringstream.bench.cpp
 libcxx/benchmarks/unordered_set_operations.bench.cpp
 libcxx/benchmarks/vector_operations.bench.cpp
-libcxx/include/CMakeLists.txt
 libcxx/include/__algorithm/binary_search.h
 libcxx/include/__algorithm/clamp.h
 libcxx/include/__algorithm/comp.h
 libcxx/include/__algorithm/comp_ref_type.h
 libcxx/include/__algorithm/copy_backward.h
 libcxx/include/__algorithm/copy.h
 libcxx/include/__algorithm/copy_if.h
 libcxx/include/__algorithm/copy_move_common.h

 libcxx/include/__concepts/regular.h
 libcxx/include/__concepts/relation.h
 libcxx/include/__concepts/same_as.h
 libcxx/include/__concepts/semiregular.h
 libcxx/include/__concepts/swappable.h
 libcxx/include/__concepts/totally_ordered.h
 libcxx/include/condition_variable
 libcxx/include/__config
-libcxx/include/__config_site.in
 libcxx/include/__coroutine/coroutine_handle.h
 libcxx/include/__coroutine/coroutine_traits.h
 libcxx/include/__coroutine/noop_coroutine_handle.h
 libcxx/include/__coroutine/trivial_awaitables.h
 libcxx/include/cstddef
 libcxx/include/ctype.h
 libcxx/include/cuchar
 libcxx/include/__debug

 libcxx/include/__iterator/readable_traits.h
 libcxx/include/__iterator/reverse_access.h
 libcxx/include/__iterator/reverse_iterator.h
 libcxx/include/__iterator/size.h
 libcxx/include/__iterator/sortable.h
 libcxx/include/__iterator/unreachable_sentinel.h
 libcxx/include/__iterator/wrap_iter.h
 libcxx/include/latch
-libcxx/include/libcxx.imp
 libcxx/include/limits
 libcxx/include/limits.h
 libcxx/include/list
 libcxx/include/__locale
 libcxx/include/locale
 libcxx/include/locale.h
 libcxx/include/map
 libcxx/include/math.h

 libcxx/include/__memory/ranges_uninitialized_algorithms.h
 libcxx/include/__memory/raw_storage_iterator.h
 libcxx/include/__memory/shared_ptr.h
 libcxx/include/__memory/temporary_buffer.h
 libcxx/include/__memory/temp_value.h
 libcxx/include/__memory/uninitialized_algorithms.h
 libcxx/include/__memory/unique_ptr.h
 libcxx/include/__memory/uses_allocator.h
-libcxx/include/module.modulemap.in
 libcxx/include/mutex
 libcxx/include/__mutex_base
 libcxx/include/new
 libcxx/include/__node_handle
 libcxx/include/numbers
 libcxx/include/numeric
 libcxx/include/__numeric/accumulate.h
 libcxx/include/__numeric/adjacent_difference.h

 libcxx/include/__utility/unreachable.h
 libcxx/include/valarray
 libcxx/include/variant
 libcxx/include/__variant/monostate.h
 libcxx/include/vector
 libcxx/include/__verbose_abort
 libcxx/include/wchar.h
 libcxx/include/wctype.h
-libcxx/src/CMakeLists.txt
 libcxx/src/any.cpp
 libcxx/src/atomic.cpp
 libcxx/src/barrier.cpp
 libcxx/src/bind.cpp
 libcxx/src/charconv.cpp
 libcxx/src/chrono.cpp
 libcxx/src/condition_variable.cpp
 libcxx/src/condition_variable_destructor.cpp

libcxx/utils/generate_ignore_format.sh

 #!/bin/bash
 
 # Generate a list of headers that aren't formatted yet
 
 if [ -z "${CLANG_FORMAT}" ]; then
   CLANG_FORMAT=clang-format
 fi
 
 rm libcxx/utils/data/ignore_format.txt
-for file in $(find libcxx/{benchmarks,include,src}/ -type f); do
+# This uses the same matches as the check-format CI step.
+#
+# Since it's hard to match empty extensions the following
+# method is used, remove all files with an extension, then
+# add the list of extensions that should be formatted.
+for file in $(find libcxx/{benchmarks,include,src}/ -type f -not -name '*.*' -or \( \
+   -name "*.h" -or -name "*.hpp" -or \
+   -name "*.c" -or -name "*.cpp" -or \
+   -name "*.inc" -or -name "*.ipp" \
+   \) ); do
+
   ${CLANG_FORMAT} --Werror --dry-run ${file} >& /dev/null
   if [ $? != 0 ]; then
     echo ${file} >> libcxx/utils/data/ignore_format.txt
   fi
 done
 
-sort libcxx/utils/data/ignore_format.txt -d -o libcxx/utils/data/ignore_format.txt
+# Force sorting not to depend on the system's locale.
+LC_ALL=C sort libcxx/utils/data/ignore_format.txt -d -o libcxx/utils/data/ignore_format.txt