mount.cifs - "Const" warnings patch

[Matthias Dieter Wallnöfer-3]

Dear mount.cifs maintainers,

we at the s4 side get always a "const" warning when compiling the "mount.cifs" tool. With this patch I substituted the existing "CONST_DISCARD" macro with the newer and better "discard_const_p" one developed by s4. When applying this one to the particular code line the warning goes away.

Matthias

[Jeff Layton-4]

On Fri, 30 Oct 2009 10:28:26 +0000 (GMT)
Matthias Dieter Wallnöfer <mdw@...> wrote:

> Dear mount.cifs maintainers,
>
> we at the s4 side get always a "const" warning when compiling the "mount.cifs" tool. With this patch I substituted the existing "CONST_DISCARD" macro with the newer and better "discard_const_p" one developed by s4. When applying this one to the particular code line the warning goes away.
>

There was no patch in this mail. Could you resend?

Thanks,
--
Jeff Layton <jlayton@...>

[Matthias Dieter Wallnöfer-3]

Hi Jeff,

sorry, sorry - sometimes I forget the attachments.

Now it should be okay.

Matthias

Jeff Layton wrote:

> On Fri, 30 Oct 2009 10:28:26 +0000 (GMT)
> Matthias Dieter Wallnöfer<mdw@...>  wrote:
>
>    
>> Dear mount.cifs maintainers,
>>
>> we at the s4 side get always a "const" warning when compiling the "mount.cifs" tool. With this patch I substituted the existing "CONST_DISCARD" macro with the newer and better "discard_const_p" one developed by s4. When applying this one to the particular code line the warning goes away.
>>
>>      
> There was no patch in this mail. Could you resend?
>
> Thanks,
>    

[Jeff Layton-4]

On Mon, 02 Nov 2009 21:36:19 +0100
Matthias Dieter Wallnöfer <mdw@...> wrote:

> Hi Jeff,
>
> sorry, sorry - sometimes I forget the attachments.
>
> Now it should be okay.
>

No problem, I do the same thing sometimes.

The patch looks safe enough, but I wonder whether the problem is really
in addmntent. Looking at a recent glibc, it doesn't look like addmntent
ever changes any of the strings it gets passed. If any of them need to
be reencoded, it uses alloca to get a new buffer and reencodes the
string into that. Maybe the prototype for it should declare the struct
as const.

Fixing it there though may be problematic. Assuming that not stripping
off the const generates a warning, maybe it would be a little more
clear to just declare a non-const variable to hold a "cifs" string and
pass that in instead. Something like this maybe?

--
Jeff Layton <jlayton@...>

---

From 13f490451f0bf536572a1f61cd0602410f8b1eff Mon Sep 17 00:00:00 2001
From: Jeff Layton <jlayton@...>
Date: Sat, 7 Nov 2009 08:15:53 -0500
Subject: [PATCH] mount.cifs: get rid of CONST_DISCARD

Apparently, we need to strip the "const" attribute off of the mnt_fstype
before passing it to addmntent to prevent a (somewhat bogus) compiler
warning.

Rather than just stripping off the "const" attribute, clarify the code
by declaring a new non-const char pointer that points to the same
string. We can also use that same pointer in the mount(2) call too.

Signed-off-by: Jeff Layton <jlayton@...>
---
 client/mount.cifs.c |    7 +++----
 1 files changed, 3 insertions(+), 4 deletions(-)

diff --git a/client/mount.cifs.c b/client/mount.cifs.c
index 3baaad7..a9c1827 100644
--- a/client/mount.cifs.c
+++ b/client/mount.cifs.c
@@ -76,8 +76,6 @@
 
 #define MAX_UNC_LEN 1024
 
-#define CONST_DISCARD(type, ptr)      ((type) ((void *) (ptr)))
-
 #ifndef SAFE_FREE
 #define SAFE_FREE(x) do { if ((x) != NULL) {free(x); x=NULL;} } while(0)
 #endif
@@ -123,6 +121,7 @@ static char * user_name = NULL;
 static char * mountpassword = NULL;
 char * domain_name = NULL;
 char * prefixpath = NULL;
+char *cifs_fstype = "cifs";
 
 /* glibc doesn't have strlcpy, strlcat. Ensure we do. JRA. We
  * don't link to libreplace so need them here. */
@@ -1590,7 +1589,7 @@ mount_retry:
  if (verboseflag)
  fprintf(stderr, "\n");
 
- if (!fakemnt && mount(dev_name, mountpoint, "cifs", flags, options)) {
+ if (!fakemnt && mount(dev_name, mountpoint, cifs_fstype, flags, options)) {
  switch (errno) {
  case ECONNREFUSED:
  case EHOSTUNREACH:
@@ -1638,7 +1637,7 @@ mount_retry:
  }
  mountent.mnt_fsname = dev_name;
  mountent.mnt_dir = mountpoint;
- mountent.mnt_type = CONST_DISCARD(char *,"cifs");
+ mountent.mnt_type = cifs_fstype;
  mountent.mnt_opts = (char *)malloc(220);
  if(mountent.mnt_opts) {
  char * mount_user = getusername();
--
1.6.0.6

---

[Matthias Dieter Wallnöfer-3]

Hi Jeff!

Yeah, please apply! My mission of this patch(set) was to prevent "CONST" warnings. Your solution does it in a very elegant manner (I would have only introduced another discard-const macro - which is far from perfect). Therefore it should be perfectly okay!

Thanks for working with me on the CONST issue.

Matthias

--- Jeff Layton <jlayton@...> schrieb am Sa, 7.11.2009:

Von: Jeff Layton <jlayton@...>
Betreff: Re: mount.cifs - "Const" warnings patch
An: "Matthias Dieter Wallnöfer" <mdw@...>
CC: "Samba Technical Mailinglist" <samba-technical@...>
Datum: Samstag, 7. November 2009, 14:19

On Mon, 02 Nov 2009 21:36:19 +0100
Matthias Dieter Wallnöfer <mdw@...> wrote:

> Hi Jeff,
>
> sorry, sorry - sometimes I forget the attachments.
>
> Now it should be okay.
>

No problem, I do the same thing sometimes.

The patch looks safe enough, but I wonder whether the problem is really
in addmntent. Looking at a recent glibc, it doesn't look like addmntent
ever changes any of the strings it gets passed. If any of them need to
be reencoded, it uses alloca to get a new buffer and reencodes the
string into that. Maybe the prototype for it should declare the struct
as const.

Fixing it there though may be problematic. Assuming that not stripping
off the const generates a warning, maybe it would be a little more
clear to just declare a non-const variable to hold a "cifs" string and
pass that in instead. Something like this maybe?

--
Jeff Layton <jlayton@...>

[Jeff Layton-4]

On Sat, 7 Nov 2009 15:45:45 +0000 (GMT)
Matthias Dieter Wallnöfer <mdw@...> wrote:

> Hi Jeff!
>
> Yeah, please apply! My mission of this patch(set) was to prevent "CONST" warnings. Your solution does it in a very elegant manner (I would have only introduced another discard-const macro - which is far from perfect). Therefore it should be perfectly okay!
>
> Thanks for working with me on the CONST issue.
>
> Matthias
>
> --- Jeff Layton <jlayton@...> schrieb am Sa, 7.11.2009:
>
> Von: Jeff Layton <jlayton@...>
> Betreff: Re: mount.cifs - "Const" warnings patch
> An: "Matthias Dieter Wallnöfer" <mdw@...>
> CC: "Samba Technical Mailinglist" <samba-technical@...>
> Datum: Samstag, 7. November 2009, 14:19
>
> On Mon, 02 Nov 2009 21:36:19 +0100
> Matthias Dieter Wallnöfer <mdw@...> wrote:
>
> > Hi Jeff,
> >
> > sorry, sorry - sometimes I forget the attachments.
> >
> > Now it should be okay.
> >
>
> No problem, I do the same thing sometimes.
>
> The patch looks safe enough, but I wonder whether the problem is really
> in addmntent. Looking at a recent glibc, it doesn't look like addmntent
> ever changes any of the strings it gets passed. If any of them need to
> be reencoded, it uses alloca to get a new buffer and reencodes the
> string into that. Maybe the prototype for it should declare the struct
> as const.
>
> Fixing it there though may be problematic. Assuming that not stripping
> off the const generates a warning, maybe it would be a little more
> clear to just declare a non-const variable to hold a "cifs" string and
> pass that in instead. Something like this maybe?
>

Pushed to master.

Matthias, do you think this needs to go to any other branches?

--
Jeff Layton <jlayton@...>

[Matthias Dieter Wallnöfer-3]

Hi Jeff,

Jeff Layton wrote:
> Pushed to master.
>
> Matthias, do you think this needs to go to any other branches?
>    
I'm not aware of any other locations.

Matthias