svn commit: r1025660 - /subversion/branches/performance/subversion/libsvn_subr/io.c

[hw...@apache.org]

Author: hwright
Date: Wed Oct 20 18:23:15 2010
New Revision: 1025660

URL: http://svn.apache.org/viewvc?rev=1025660&view=rev
Log:
On the performance branch:

* subversion/libsvn_subr/io.c:
  After on-list discussion, revert parts of r985472.  See this post for
  further enlightenment:
    http://svn.haxx.se/dev/archive-2010-10/0257.shtml

  The crux is that callers which don't care about specific names have an
  API to use for that matter, so we shouldn't be introducing additional
  randomness here.

Modified:
    subversion/branches/performance/subversion/libsvn_subr/io.c

Modified: subversion/branches/performance/subversion/libsvn_subr/io.c
URL: http://svn.apache.org/viewvc/subversion/branches/performance/subversion/libsvn_subr/io.c?rev=1025660&r1=1025659&r2=1025660&view=diff
==============================================================================
--- subversion/branches/performance/subversion/libsvn_subr/io.c (original)
+++ subversion/branches/performance/subversion/libsvn_subr/io.c Wed Oct 20 18:23:15 2010
@@ -378,11 +378,6 @@ svn_io_open_uniquely_named(apr_file_t **
       if (delete_when == svn_io_file_del_on_close)
         flag |= APR_DELONCLOSE;
 
-      /* Increase the chance that rand() will return something truely
-         independent from what others get or do. */
-      if (i == 2)
-        srand((unsigned int)apr_time_now());
-
       /* Special case the first attempt -- if we can avoid having a
          generated numeric portion at all, that's best.  So first we
          try with just the suffix; then future tries add a number
@@ -393,16 +388,11 @@ svn_io_open_uniquely_named(apr_file_t **
          This is good, since "1" would misleadingly imply that
          the second attempt was actually the first... and if someone's
          got conflicts on their conflicts, we probably don't want to
-         add to their confusion :-). 
-
-         Also, the randomization used to minimize the number of re-try 
-         cycles will interfere with certain tests that compare working
-         copies etc.
-       */
+         add to their confusion :-). */
       if (i == 1)
         unique_name = apr_psprintf(scratch_pool, "%s%s", path, suffix);
       else
-        unique_name = apr_psprintf(scratch_pool, "%s.%u_%x%s", path, i, rand(), suffix);
+        unique_name = apr_psprintf(scratch_pool, "%s.%u_%s", path, i, suffix);
 
       /* Hmmm.  Ideally, we would append to a native-encoding buf
          before starting iteration, then convert back to UTF-8 for