On Wednesday, August 4, 2021 9:59:30 AM CEST Dan Carpenter wrote: > On Fri, Jul 30, 2021 at 08:14:52PM +0200, Fabio M. De Francesco wrote: > > static inline void __nat25_generate_ipx_network_addr_with_socket(unsigned char > > *networkAddr,> > > - unsigned int *ipxNetAddr, unsigned short *ipxSocketAddr) > > > > [...] > > > Here is another bug which was obscured/caused by the union. > > addr.f0 = be16_to_cpu(*ipxSocketAddr); > > The addr.f0 variable is an int. On big endian systems only the last two > bytes are set: > > memcpy(networkAddr+5, addr.f1, 2); > > So this is the equivalent of: > > memset(networkAddr+5, 0, 2); > > regards, > dan carpenter Dear Dan, Thanks, for pointing me to one more bug I introduced with this patch. The most of them were due to me forgetting that memcpy() takes pointers. For some reason I was thinking it takes values, therefore I put in it a lot of unnecessary and faulty complications. I'd like to make a new patch, a better one (I hope), without unneeded unions without the other wrong lines that are in the commit 56febcc2595e. However, I see that Greg hasn't yet had the time to revert the above commit, so I don't know how to make a new patch. I mean: I could (1) either wait for Greg to revert it and then to fix the sparse warnings with a new patch, or (2) I could fix the bugs I made in 56febcc2595e without having it reverted. I would prefer the solution (2) with a "Fixes: 56febcc2595e (...)" and a "Reported-by: Dan Carpenter <...>" tags. What is the best solution between the two above? Thanks, Fabio
