Hi,
Just a couple of liner fix, would appreciate if someone can take a quick
look.
Bug description:
http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6300049
Webrev:
http://cr.opensolaris.org/~shri.k/webrev_6300049/
The bug isn't really: "netperf -f dst" works fine for 16 bit prefix but
fails for 48 bit prefix.
"netstat -f dst" *may* fail for any IPv6 prefix containing byte value >=
128 because:
- Function v6_addr_match:4311 compares *ucp (uint8_t value) with *aptr
(char value).
- uint8_t has range 0-255, so *ucp is fine, but
- for *aptr: C standard allows char to default to either signed (-128 to
127) or unsigned (0 to 255) while leaving that choice to the compiler
implementation/platform.
Thus, the platforms that defaults char to signed char would run into
this bug. Interpreting the value as unsigned 8 bit integer and then
comparing with *ucp makes the implementation platform independent. (I
have tested the fix, it works fine).
While this fix will resolve the bug, am not sure if we have more of such
bugs (as yet undetected) esp. since struct hostent stores the address in
char **h_addr_list.
Thanks,
Shri
_______________________________________________
networking-discuss mailing list
[email protected]
