Port better to NFS unlink

I found this problem while looking into Bug#72641.
* lib-src/etags.c (do_move_file):
* lib-src/update-game-score.c (unlock_file):
* src/androidvfs.c (android_hack_asset_fd_fallback):
* src/filelock.c (current_lock_owner):
Treat unlink as successful if it fails because the file wasn’t there.
This can happen with some NFS implementations, due to its
retrying over the network to get at-least-once semantics.
Although most of Emacs’s calls to unlink were already doing this,
a few instances were not.

2 files changed, 2 insertions, 2 deletions

diff --git a/src/androidvfs.c b/src/androidvfs.c
index 14da8eed37e..ff81ef288f5 100644
--- a/src/androidvfs.c
+++ b/src/androidvfs.c
@@ -1323,7 +1323,7 @@ android_hack_asset_fd_fallback (AAsset *asset)
   if (fd < 0)
     return -1;
 
-  if (unlink (filename))
+  if (unlink (filename) && errno != ENOENT)
     goto fail;
 
   if (ftruncate (fd, size))
diff --git a/src/filelock.c b/src/filelock.c
index 1ae57dc7344..bc09fce69f8 100644
--- a/src/filelock.c
+++ b/src/filelock.c
@@ -501,7 +501,7 @@ current_lock_owner (lock_info_type *owner, Lisp_Object lfname)
      the file system is buggy, e.g., <https://bugs.gnu.org/72641>.
      Emacs never creates empty lock files even temporarily, so removing
      an empty lock file should be harmless.  */
-  return emacs_unlink (SSDATA (lfname)) < 0 ? errno : 0;
+  return emacs_unlink (SSDATA (lfname)) < 0 && errno != ENOENT ? errno : 0;
 }