[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [lvm-devel] Avoid more warnings; general const-correctness improvements



Jim Meyering <jim meyering net> wrote:
> Building with this
>
>     ./configure --disable-nls --with-clvmd=cman && make
>
> and the patch below (along with the others I've sent today),
> there are no more warnings.
>
> [At first, I was configuring --with-clvmd=all and made
>  a few changes for gulm-related code, but then noticed
>  it's "old", so quit looking at that part. ]
>
> Parts of the patch below may be a tiny bit controversial, since they
> work around the fact that tcp_remove_client and alloc_client both
> modify their "char *csid" parameter, while logically they needn't
> (and shouldn't) do that.  I.e., the sole alloc_client caller certainly
> doesn't expect its addr.sin6_addr argument to be modified.  Before this
> change, tcp_remove_client would modify its csid parameter, but restore
> the original value right afterwards.  The only objection I can imagine
> is that this change adds the overhead of the added memcpy calls.  IMHO,
> they're worth the price to make the interfaces cleaner and avoid warnings.
...
> 	* daemons/clvmd/tcp-comms.c (tcp_remove_client): Make parameter "const".
> 	But since it was being modified, use a temporary copy and modify that
> 	instead.  Use a loop rather than two identical blocks of code.
> 	(alloc_client): Similarly, make the "csid" parameter "const",
> 	and modify a copy.

FYI, I've just noticed that tcp-comms.c (and hence the potentially
objectionable hunks) is used only --with-clvmd=gulm (or =all),
so I suppose it can be omitted, if it's an issue.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]