RE: Windows CVS 1.12.12.1 Latest + Patch - "version" Command Hangs But :pserver: Not Busy

[Conrad T. Pino]

> From: Derek Price [mailto:derek@ximbiot.com]
> 
> >     need    4294967286
> 
> That does look awfully big.  Some sort of initialization or byte-order
> problem, perhaps?  Byte-order sounds unlikely since this runs on x86
> linux, but figuring out where that value came from should provide a clue.

I think it's an underflow bug since as both "need" and "size" are "size_t"
which is "unsigned int" for VC6.

If "nbytes" > "need" then "need -= nbytes;" makes "need" *BIG* quickly.

A patch the like following will help but IMO the real questions are:

1. Did "need" change type from "int" to "size_t" recently?

2. Why is the server returning more than "need"?

3. Is the initial "need" smaller than it used to be?

4. Where else might we have similar underflow problems?

Index: src/socket-client.c
===================================================================
RCS file: /cvsroot/cvs/ccvs/src/socket-client.c,v
retrieving revision 1.15
diff -u -p -r1.15 socket-client.c
--- src/socket-client.c 26 May 2005 08:24:17 -0000      1.15
+++ src/socket-client.c 24 Sep 2005 22:15:02 -0000
@@ -143,8 +143,16 @@ socket_buffer_input( void *closure, char
            else
                return 0;
        }
-       need -= nbytes;
-       size -= nbytes;
+       if (nbytes < need)
+               need -= nbytes;
+       else
+               need = 0;
+
+       if (nbytes < size)
+               size -= nbytes;
+       else
+               size = 0;
+
        data += nbytes;
        *got += nbytes;
     }