This is the mail archive of the cygwin-patches mailing list for the Cygwin 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]

Try to remove possible race pinfo::init


On some systems, I was frequent segmentation faults on fork(), and
I was able to track it down to the patch I just sent. However, my
scripts don't run much better, because they now fail only a bit
more cleanly with "fork: Resource temporarily unavailable".

Since my previous patch transformed seg faults to EAGAIN errors,
I tried to find potential races or other errors in pinfo::init.
I didn't really take the time to try to understand the code, but
if I'm guessing right, the MapViewOfFileEx() call is doing something
that depends on the child. If that's right, then it seems to me
that the retry loop in open_shared() failure case is a bit too
tight.


2005-07-29  Arto Huusko  <arto.huusko@wmdata.fi>

	* pinfo.cc (pinfo::init): Sleep before retrying open_shared().

--- pinfo.cc	2005-07-29 11:25:43.501687460 +0300
+++ pinfo.cc.fixed	2005-07-29 11:26:17.931478519 +0300
@@ -235,6 +235,7 @@ pinfo::init (pid_t n, DWORD flag, HANDLE
 	    }
 
 	  debug_printf ("MapViewOfFileEx h0 %p, i %d failed, %E", h0, i);
+	  low_priority_sleep (0);
 	  continue;
 	}
 

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