1 files changed, 41 insertions, 0 deletions
diff --git a/mit-pthreads/stdio/README b/mit-pthreads/stdio/README
new file mode 100755
index 00000000000..8e4c07909e9
--- /dev/null
+++ b/mit-pthreads/stdio/README
@@ -0,0 +1,41 @@
+Copyright (c) 1993, 1994 Chris Provenzano. All rights reserved.
+
+This is a threadsafe stdio based on the BSD stdio written by Chris Torek.
+This product includes software developed by the Univeristy of California,
+Berkeley and its contributors.
+
+INCLUDE FILES AND PORTING
+To continue to make this package portable, some basic rules on includes
+files must be followed.
+
+pthread.h should be included first (if it is to be included).
+stdio.h should be included.
+
+INTERNAL LOCKING
+1.	All functions that can be called by the user must have flockfile() at the
+	begining and a funlockfile() at the end. The routine flockfile() is a
+	counting mutex, The thread that owns the lock can call flockfile() as 
+	many times as it wants, but must call an equal number of funlockfile()
+	before the lock will be released.
+2.	All functions starting with __ shouldn't need addtional locking.
+3.	Anything that writes the variable __sglue should lock __sfp_mutex,
+ 	check __sfp_state, and do a condion wait if it is set.
+4.	Anything that checks fp->_flag for valididity should also lock
+	__sfp_mutex.
+5.	Anything that reads the variable __sglue should lock __sfp_mutex, increment
+	__sfp_state, and then unlock the mutex. At function return it should
+	lock the mutex again decrement __sfp_state and check if zero. If so
+	do a cond_signal, and unlock the mutex.
+6.	The functions fopen, fdopen, and freopen are the only functions that
+	will change a fp->_file
+7.	fdopen and fopen both allocate the next fp by locking __sfp_mutex
+	checking fp->_flags and then setting it if free.
+8.	freopen tries to preserve fp->_file. It sets __sfp_mutex, then it
+	tries to lock fp->_file and close it.
+9.	__sinit is done with a pthread_once routine.
+
+
+
+Things to do.
+
+Fix printf so it uses the ininf function.