On Jan 18 13:40, Ben wrote:
> On 18-01-2021 11:56, Corinna Vinschen via Cygwin-patches wrote:
> > Hmm, you're adding another function call to the call stack.  Doesn't
> > that slow down _remove_r rather than speeding it up?  Ok, this function
> > is called from _tmpfile_r/_tmpfile64_r only, so dedup may trump speed
> > here...
> > 
> > What's your stance?
> > 
> While I could do without:
> In an earlier version I had changed remove and missed remove_r.
> 
> So, this commit is more about de-duplication rather than speed.

What about this instead?  It should be better optimizable:

diff --git a/winsup/cygwin/syscalls.cc b/winsup/cygwin/syscalls.cc
index 4742c665339c..2d8acb4c1052 100644
--- a/winsup/cygwin/syscalls.cc
+++ b/winsup/cygwin/syscalls.cc
@@ -1133,24 +1133,15 @@ _remove_r (struct _reent *, const char *ourname)
       return -1;
     }
 
-  return win32_name.isdir () ? rmdir (ourname) : unlink (ourname);
+  int res = win32_name.isdir () ? rmdir (ourname) : unlink (ourname);
+  syscall_printf ("%R = remove(%s)", res, ourname);
+  return res;
 }
 
 extern "C" int
 remove (const char *ourname)
 {
-  path_conv win32_name (ourname, PC_SYM_NOFOLLOW);
-
-  if (win32_name.error)
-    {
-      set_errno (win32_name.error);
-      syscall_printf ("-1 = remove (%s)", ourname);
-      return -1;
-    }
-
-  int res = win32_name.isdir () ? rmdir (ourname) : unlink (ourname);
-  syscall_printf ("%R = remove(%s)", res, ourname);
-  return res;
+  return _remove_r (_GLOBAL_REENT, ourname);
 }
 
 extern "C" pid_t



Corinna

Reply via email to