[PATCH][ATM]: [lanai] ioctl only meant for debugging (from mitch@sfgoth.com)

[PATCH][ATM]: [lanai] ioctl only meant for debugging (from mitch@sfgoth.com)

[chas williams]

pleae apply to 2.4. --thanks

# This is a BitKeeper generated patch for the following project:
# Project Name: Linux kernel tree
# This patch format is intended for GNU patch command version 2.5 or higher.
# This patch includes the following deltas:
#	           ChangeSet	1.1115  -> 1.1116 
#	 drivers/atm/lanai.c	1.4     -> 1.5    
#
# The following is the BitKeeper ChangeSet Log
# --------------------------------------------
# 03/09/22	chas@relax.cmf.nrl.navy.mil	1.1116
# [ATM]: [lanai] ioctl only meant for debugging (from mitch@sfgoth.com)
# --------------------------------------------
#
diff -Nru a/drivers/atm/lanai.c b/drivers/atm/lanai.c
--- a/drivers/atm/lanai.c	Mon Sep 22 13:01:23 2003
+++ b/drivers/atm/lanai.c	Mon Sep 22 13:01:23 2003
@@ -2574,6 +2574,7 @@
 /* NOTE: these are all DEBUGGING ONLY currently */
 static int lanai_ioctl(struct atm_dev *atmdev, unsigned int cmd, void *arg)
 {
+#if 0
 	int result = 0;
 	struct lanai_dev *lanai = (struct lanai_dev *) atmdev->dev_data;
 	switch(cmd) {
@@ -2653,6 +2654,10 @@
 			result = -EINVAL;
 	}
 	return result;
+#else /* !0 */
+	(void) atmdev; (void) cmd; (void) arg;  /* no compiler warnings */
+	return -ENOIOCTLCMD;
+#endif /* 0 */
 }
 
 static int lanai_send(struct atm_vcc *atmvcc, struct sk_buff *skb)

Re: [PATCH][ATM]: [lanai] ioctl only meant for debugging (from mitch@sfgoth.com)

[Christoph Hellwig]

Umm, shouldn't you just ifdef out the whole function and it's
assignment to the operations vector?  This looks horribly ugly..

Re: [PATCH][ATM]: [lanai] ioctl only meant for debugging (from mitch@sfgoth.com)

[chas williams]

In message <20030922180749.A26432@infradead.org>,Christoph Hellwig writes:
>Umm, shouldn't you just ifdef out the whole function and it's
>assignment to the operations vector?  This looks horribly ugly..

i suppose i am not convinced that two #ifdef/#endif's are better than an
#if/#else/#endif

Re: [PATCH][ATM]: [lanai] ioctl only meant for debugging (from mitch@sfgoth.com)

[Christoph Hellwig]

On Mon, Sep 22, 2003 at 01:55:07PM -0400, chas williams wrote:
> In message <20030922180749.A26432@infradead.org>,Christoph Hellwig writes:
> >Umm, shouldn't you just ifdef out the whole function and it's
> >assignment to the operations vector?  This looks horribly ugly..
> 
> i suppose i am not convinced that two #ifdef/#endif's are better than an
> #if/#else/#endif

first ifdefs in a function are always worse than around functions,
second ifdefs in a function that make it a stuv are pretty ugly and
third your (void)arg crap is ugly as hell. (what compiler do you have
that complains about this, btw, gcc 3.3 doesn't..).

So if you prefer one if/else/endif you can do it as

#if FOO
foo_ioctl()
{
}
#else
#define foo_ioctl	NULL
#endif

but in the case of assigning a function pointer to an operation vector
I find this more confusing than the above variant. 

---end quoted text---

Re: [PATCH][ATM]: [lanai] ioctl only meant for debugging (from mitch@sfgoth.com)

[Mitchell Blank Jr]

Christoph Hellwig wrote (in two different emails):
> Umm, shouldn't you just ifdef out the whole function and it's
> assignment to the operations vector?  This looks horribly ugly..

Sorry for any ugliness - its my fault.  I personally don't think it really
makes a difference either way.  Its already #ifdef'ed the same way in 2.6.

> and
> third your (void)arg crap is ugly as hell. (what compiler do you have
> that complains about this, btw, gcc 3.3 doesn't..).

"gcc -Wunused-parameter"

It's true that the kernel doesn't compile with that option but I've gotten
in the habit of adding the "(void) x;" from working on other projects that
do.  I think it's pretty harmless.

> #if FOO
> foo_ioctl()
> {
> }
> #else
> #define foo_ioctl	NULL
> #endif

Yes, that would probably be (a bit) cleaner.  That's what we already
use for lanai_proc_read if !CONFIG_PROC_FS

-Mitch

Re: [PATCH][ATM]: [lanai] ioctl only meant for debugging (from mitch@sfgoth.com)

[David S. Miller]

On Mon, 22 Sep 2003 13:55:07 -0400
chas williams <chas@cmf.nrl.navy.mil> wrote:

> i suppose i am not convinced that two #ifdef/#endif's are better than an
> #if/#else/#endif

True, but how about we use an LANAI_DEBUG or something instead of "#if
0" that can sit undef'd at the top of the file?

Re: [PATCH][ATM]: [lanai] ioctl only meant for debugging (from mitch@sfgoth.com)

[chas williams]

In message <20030923035729.2aa1afd6.davem@redhat.com>,"David S. Miller" writes:
>On Mon, 22 Sep 2003 13:55:07 -0400
>chas williams <chas@cmf.nrl.navy.mil> wrote:
>
>> i suppose i am not convinced that two #ifdef/#endif's are better than an
>> #if/#else/#endif
>
>True, but how about we use an LANAI_DEBUG or something instead of "#if
>0" that can sit undef'd at the top of the file?

i suppose i might suggest DEBUG_RW which already exists in the code.

/*
 * Debug _all_ register operations with card, except the memory test.
 * Also disables the timed poll to prevent extra chattiness.  This
 * isn't for normal use
 */
#undef DEBUG_RW

DEBUG seems to be typically defined so isnt a great choice.  any
comments mitch?