This is the mail archive of the libstdc++@gcc.gnu.org mailing list for the libstdc++ project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[Patch] Remove a dead seekoff from _M_underflow


Hi,

consider this code from the current _M_underflow:

if (this->_M_in_cur < this->_M_in_end)
  {
     __ret = traits_type::to_int_type(*this->_M_in_cur);
     if (__bump)
       _M_move_in_cur(1);
     return __ret;
  }

// Sync internal and external buffers.
// NB: __testget -> __testput as _M_buf_unified here.
if (this->_M_in_cur > this->_M_in_beg)
  {
    if (__testout)
      _M_overflow();
    else if (this->_M_in_cur != _M_filepos)
      _M_file.seekoff(this->_M_in_cur - _M_filepos, ios_base::cur,
                      ios_base::in);
  }

When !__testout for sure _M_out_cur_move is never called and this
implies that _M_in_end is stuck at the value set by the last
_M_set_buffer call. Since in the code above __testin is true,
_M_in_end == _M_filepos (see _M_set_buffer) and therefore, in the
second block above always _M_in_cur == _M_filepos.

I also take the occasion to rewrite the resulting simplified 'if'
in terms of _M_out_* pointers (equivalent but more clear, IMO) and
remove a redundant 'if' not shown above: everywhere else
_M_destroy_pback is called without checking for _M_pback_init, this
is only a remnant from a previous more convoluted 'if' block.

Tested x86-linux, also put an abort() in the original code in place
of the removed 'if' branch and ran the testsuite.

Paolo.

//////////
2003-05-22  Paolo Carlini  <pcarlini@unitus.it>

	* include/bits/fstream.tcc (_M_underflow): simplify:
	!__testout implies _M_filepos == _M_in_end, therefore
	the first _M_file.seekoff call is never issued.
diff -urN libstdc++-v3-orig/include/bits/fstream.tcc libstdc++-v3/include/bits/fstream.tcc
--- libstdc++-v3-orig/include/bits/fstream.tcc	2003-05-16 01:43:14.000000000 +0200
+++ libstdc++-v3/include/bits/fstream.tcc	2003-05-22 13:05:45.000000000 +0200
@@ -196,8 +196,7 @@
 	  // Check for pback madness, and if so swich back to the
 	  // normal buffers and jet outta here before expensive
 	  // fileops happen...
-	  if (_M_pback_init)
-	    _M_destroy_pback();
+	  _M_destroy_pback();
 
 	  if (this->_M_in_cur < this->_M_in_end)
 	    {
@@ -208,15 +207,8 @@
 	    }
 
 	  // Sync internal and external buffers.
-	  // NB: __testget -> __testput as _M_buf_unified here.
-	  if (this->_M_in_cur > this->_M_in_beg)
-	    {
-	      if (__testout)
-		_M_overflow();
-	      else if (this->_M_in_cur != _M_filepos)
-		_M_file.seekoff(this->_M_in_cur - _M_filepos, ios_base::cur, 
-				ios_base::in);
-	    }
+	  if (__testout && this->_M_out_beg < this->_M_out_lim)
+	    _M_overflow();
 
 	  if (_M_buf_size > 1)
 	    {

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]