[18/N] [libcxx] Use the posix code for directory_entry::__do_refresh

[mstorsjo]

Summary

This works just fine for windows, as all the functions it calls are implemented and wrapped for windows.

[amccarth]

LGTM.

[amccarth]

Oh, are there tests that need to be enabled on Windows?

[mstorsjo]

In D91173#2438006, @amccarth wrote:

Oh, are there tests that need to be enabled on Windows?

There's a whole bunch of tests, and none of them are excluded on windows right now (as std::filesystem isn't possible to even build for windows as things are in git so far), so this patch reduces the amount of failing tests.

(Locally, I have a branch with a whole lot of patches on top of the testsuite for std::filesystem to disable/change bits to make it pass with MS STL, and with this patchset for the implementation in libc++, it gets to a point where the same tests that succeed with MS STL also succeed with libc++. I started out by upstreaming the least hacky bits of those patches, but there's just a whole lot of it, and it's more productive for now to focus on upstreaming the implementation. I was hoping for getting it in by the time of the llvm 12 branch, but we'll see how that goes...)

[mstorsjo]

This one already has a green report from CI back when it was uploaded, and I think the surrounding area hasn't changed since, but I can push it through CI once before pushing the code, if it's otherwise ok. This is essentially only a deletion of a windows specific ifdef block, nothing more.

[ldionne]

I love red diffs. LGTM.

As @amccarth said - I'm relying on you running the tests on Windows locally for the correctness. Once we're trough with your patchset and filesystem builds on windows, I would be really really thankful if we could add CI on a windows machine and run the tests there.

libcxx/src/filesystem/operations.cpp

 }
 #endif
 
 
 ///////////////////////////////////////////////////////////////////////////////
 //                           directory entry definitions
 ///////////////////////////////////////////////////////////////////////////////
 
-#ifndef _LIBCPP_WIN32API
 error_code directory_entry::__do_refresh() noexcept {
   __data_.__reset();
   error_code failure_ec;
 
   StatT full_st;
   file_status st = detail::posix_lstat(__p_, full_st, &failure_ec);
   if (!status_known(st)) {
     __data_.__reset();
if (_VSTD_FS::exists(st)) {
     // the value is actually used.
     error_code ignored_ec;
     __data_.__write_time_ =
         __extract_last_write_time(__p_, full_st, &ignored_ec);
   }
 
   return failure_ec;
 }
-#else
-error_code directory_entry::__do_refresh() noexcept {
-  __data_.__reset();
-  error_code failure_ec;
-
-  file_status st = _VSTD_FS::symlink_status(__p_, failure_ec);
-  if (!status_known(st)) {
-    __data_.__reset();
-    return failure_ec;
-  }
-
-  if (!_VSTD_FS::exists(st) || !_VSTD_FS::is_symlink(st)) {
-    __data_.__cache_type_ = directory_entry::_RefreshNonSymlink;
-    __data_.__type_ = st.type();
-    __data_.__non_sym_perms_ = st.permissions();
-  } else { // we have a symlink
-    __data_.__sym_perms_ = st.permissions();
-    // Get the information about the linked entity.
-    // Ignore errors from stat, since we don't want errors regarding symlink
-    // resolution to be reported to the user.
-    error_code ignored_ec;
-    st = _VSTD_FS::status(__p_, ignored_ec);
-
-    __data_.__type_ = st.type();
-    __data_.__non_sym_perms_ = st.permissions();
-
-    // If we failed to resolve the link, then only partially populate the
-    // cache.
-    if (!status_known(st)) {
-      __data_.__cache_type_ = directory_entry::_RefreshSymlinkUnresolved;
-      return error_code{};
-    }
-    __data_.__cache_type_ = directory_entry::_RefreshSymlink;
-  }
-
-  // FIXME: This is currently broken, and the implementation only a placeholder.
-  // We need to cache last_write_time, file_size, and hard_link_count here before
-  // the implementation actually works.
-
-  return failure_ec;
-}
-#endif
 
 _LIBCPP_END_NAMESPACE_FILESYSTEM