EXPORT_SYMBOL generates "is deprecated" noise

EXPORT_SYMBOL generates "is deprecated" noise

[Martin J. Bligh]

I'm getting lots of errors like this nowadays:

drivers/serial/8250.c:2651: warning: `register_serial' is deprecated 
(declared at drivers/serial/8250.c:2607)

Which is just: "EXPORT_SYMBOL(register_serial);"

Sorry, but that's just compile-time noise, not anything useful.
Warning on real usages of it might be handy (we can go fix the users)
but not EXPORT_SYMBOL - we can't kill the export until the function
goes away. The more noise we have, the harder it is to see real errors 
and warnings.

I took a quick poke around, but can't see what generates this stuff.
What is doing these checks, and can we please make an exception for
EXPORT_SYMBOL (and EXPORT_SYMBOL_GPL) somehow?

M.

Re: EXPORT_SYMBOL generates "is deprecated" noise

[Adrian Bunk]

On Sun, Aug 07, 2005 at 11:07:59AM -0700, Martin J. Bligh wrote:
> I'm getting lots of errors like this nowadays:
> 
> drivers/serial/8250.c:2651: warning: `register_serial' is deprecated 
> (declared at drivers/serial/8250.c:2607)
> 
> Which is just: "EXPORT_SYMBOL(register_serial);"
> 
> Sorry, but that's just compile-time noise, not anything useful.
> Warning on real usages of it might be handy (we can go fix the users)
> but not EXPORT_SYMBOL - we can't kill the export until the function
> goes away. The more noise we have, the harder it is to see real errors 
> and warnings.
> 
> I took a quick poke around, but can't see what generates this stuff.
> What is doing these checks, and can we please make an exception for
> EXPORT_SYMBOL (and EXPORT_SYMBOL_GPL) somehow?

This is generated by the __deprecated marker (#define'd to 
__attribute__((deprecated)) ) at the prototype in 
include/linux/serial.h.

You could somehow #ifdef the warning away, but IMHO this would be more 
ugly than living with the warning until the last user is gone.

> M.

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: EXPORT_SYMBOL generates "is deprecated" noise

[Martin J. Bligh]

--Adrian Bunk <bunk@stusta.de> wrote (on Sunday, August 07, 2005 20:23:12 +0200):

> On Sun, Aug 07, 2005 at 11:07:59AM -0700, Martin J. Bligh wrote:
>> I'm getting lots of errors like this nowadays:
>> 
>> drivers/serial/8250.c:2651: warning: `register_serial' is deprecated 
>> (declared at drivers/serial/8250.c:2607)
>> 
>> Which is just: "EXPORT_SYMBOL(register_serial);"
>> 
>> Sorry, but that's just compile-time noise, not anything useful.
>> Warning on real usages of it might be handy (we can go fix the users)
>> but not EXPORT_SYMBOL - we can't kill the export until the function
>> goes away. The more noise we have, the harder it is to see real errors 
>> and warnings.
>> 
>> I took a quick poke around, but can't see what generates this stuff.
>> What is doing these checks, and can we please make an exception for
>> EXPORT_SYMBOL (and EXPORT_SYMBOL_GPL) somehow?
> 
> This is generated by the __deprecated marker (#define'd to 
> __attribute__((deprecated)) ) at the prototype in 
> include/linux/serial.h.
> 
> You could somehow #ifdef the warning away, but IMHO this would be more 
> ugly than living with the warning until the last user is gone.

If it's going to spout crap when I'm not even using the deprecated
function, it's worse than useless, I'm afraid. 

I tried some "smart" nested definitions to just wrap it inside EXPORT_SYMBOL, 
but the ordering of evaluation is all wrong. Bah ;-(

M.

Re: EXPORT_SYMBOL generates "is deprecated" noise

[Adrian Bunk]

On Sun, Aug 07, 2005 at 11:55:16AM -0700, Martin J. Bligh wrote:
> 
> 
> --Adrian Bunk <bunk@stusta.de> wrote (on Sunday, August 07, 2005 20:23:12 +0200):
> 
> > On Sun, Aug 07, 2005 at 11:07:59AM -0700, Martin J. Bligh wrote:
> >> I'm getting lots of errors like this nowadays:
> >> 
> >> drivers/serial/8250.c:2651: warning: `register_serial' is deprecated 
> >> (declared at drivers/serial/8250.c:2607)
> >> 
> >> Which is just: "EXPORT_SYMBOL(register_serial);"
> >> 
> >> Sorry, but that's just compile-time noise, not anything useful.
> >> Warning on real usages of it might be handy (we can go fix the users)
> >> but not EXPORT_SYMBOL - we can't kill the export until the function
> >> goes away. The more noise we have, the harder it is to see real errors 
> >> and warnings.
> >> 
> >> I took a quick poke around, but can't see what generates this stuff.
> >> What is doing these checks, and can we please make an exception for
> >> EXPORT_SYMBOL (and EXPORT_SYMBOL_GPL) somehow?
> > 
> > This is generated by the __deprecated marker (#define'd to 
> > __attribute__((deprecated)) ) at the prototype in 
> > include/linux/serial.h.
> > 
> > You could somehow #ifdef the warning away, but IMHO this would be more 
> > ugly than living with the warning until the last user is gone.
> 
> If it's going to spout crap when I'm not even using the deprecated
> function, it's worse than useless, I'm afraid. 
>...

It's reminding us that we are still offering a deprecated function. ;-)

> M.

cu
Adrian

-- 

       "Is there not promise of rain?" Ling Tan asked suddenly out
        of the darkness. There had been need of rain for many days.
       "Only a promise," Lao Er said.
                                       Pearl S. Buck - Dragon Seed

Re: EXPORT_SYMBOL generates "is deprecated" noise

[Martin J. Bligh]

>> If it's going to spout crap when I'm not even using the deprecated
>> function, it's worse than useless, I'm afraid. 
>> ...
> 
> It's reminding us that we are still offering a deprecated function. ;-)

Might be useful as an option. But not to irritate every poor sod who
does a kernel compile, ever.

M.

Re: EXPORT_SYMBOL generates "is deprecated" noise

[Martin J. Bligh]

--"Martin J. Bligh" <mbligh@mbligh.org> wrote (on Sunday, August 07, 2005 11:07:59 -0700):

> I'm getting lots of errors like this nowadays:
> 
> drivers/serial/8250.c:2651: warning: `register_serial' is deprecated 
> (declared at drivers/serial/8250.c:2607)
> 
> Which is just: "EXPORT_SYMBOL(register_serial);"
> 
> Sorry, but that's just compile-time noise, not anything useful.
> Warning on real usages of it might be handy (we can go fix the users)
> but not EXPORT_SYMBOL - we can't kill the export until the function
> goes away. The more noise we have, the harder it is to see real errors 
> and warnings.
> 
> I took a quick poke around, but can't see what generates this stuff.
> What is doing these checks, and can we please make an exception for
> EXPORT_SYMBOL (and EXPORT_SYMBOL_GPL) somehow?

Oh, I'm being an idiot and looking at the wrong tree. It's __deprecated,
but I still can't think of a clean way to locally undefine that for
just EXPORT_SYMBOL.

M.

Re: EXPORT_SYMBOL generates "is deprecated" noise

[Arnd Bergmann]

On Sünndag 07 August 2005 20:26, Martin J. Bligh wrote:
> Oh, I'm being an idiot and looking at the wrong tree. It's __deprecated,
> but I still can't think of a clean way to locally undefine that for
> just EXPORT_SYMBOL.

We could in theory create a new EXPORT_SYMBOL variant that does not
reference the symbol directly. This does a little less compile-time
checks but helps reduce the noise. The big advantage of this
would be that we could once again build kernels with -Werror on
developer machines.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>

diff --git a/include/linux/module.h b/include/linux/module.h
--- a/include/linux/module.h
+++ b/include/linux/module.h
@@ -182,21 +182,26 @@ void *__symbol_get_gpl(const char *symbo
 #endif
 
 /* For every exported symbol, place a struct in the __ksymtab section */
-#define __EXPORT_SYMBOL(sym, sec)				\
-	__CRC_SYMBOL(sym, sec)					\
-	static const char __kstrtab_##sym[]			\
+#define __EXPORT_SYMBOL(name, sym, sec)				\
+	__CRC_SYMBOL(name, sec)					\
+	static const char __kstrtab_##name[]			\
 	__attribute__((section("__ksymtab_strings")))		\
-	= MODULE_SYMBOL_PREFIX #sym;                    	\
-	static const struct kernel_symbol __ksymtab_##sym	\
+	= MODULE_SYMBOL_PREFIX #name;                    	\
+	static const struct kernel_symbol __ksymtab_##name	\
 	__attribute_used__					\
 	__attribute__((section("__ksymtab" sec), unused))	\
-	= { (unsigned long)&sym, __kstrtab_##sym }
+	= { (unsigned long)&sym, __kstrtab_##name }
 
 #define EXPORT_SYMBOL(sym)					\
-	__EXPORT_SYMBOL(sym, "")
+	__EXPORT_SYMBOL(sym, sym, "")
 
 #define EXPORT_SYMBOL_GPL(sym)					\
-	__EXPORT_SYMBOL(sym, "_gpl")
+	__EXPORT_SYMBOL(sym, sym, "_gpl")
+
+#define EXPORT_DEPRECATED_SYMBOL(sym)				\
+	extern void __deprecated_ ## sym 			\
+			__attribute__((alias(#sym)));		\
+	__EXPORT_SYMBOL(sym, __deprecated_ ## sym, "_gpl")
 
 #endif
 

Re: EXPORT_SYMBOL generates "is deprecated" noise

[Russell King]

On Sun, Aug 07, 2005 at 11:07:59AM -0700, Martin J. Bligh wrote:
> I'm getting lots of errors like this nowadays:
> 
> drivers/serial/8250.c:2651: warning: `register_serial' is deprecated 
> (declared at drivers/serial/8250.c:2607)
> 
> Which is just: "EXPORT_SYMBOL(register_serial);"
> 
> Sorry, but that's just compile-time noise, not anything useful.
> Warning on real usages of it might be handy (we can go fix the users)
> but not EXPORT_SYMBOL - we can't kill the export until the function
> goes away. The more noise we have, the harder it is to see real errors 
> and warnings.

I don't know why I bother with __deprecated - I haven't seen much
evidence of the users of these functions being cleaned up, so I
think we might as well just delete the functions and _force_ people
to fix their code.  That unfortunately seems to be the only way to
get things done in this day and age, which is rather sad if that's
what it takes to kick people into action.

-- 
Russell King
 Linux kernel    2.6 ARM Linux   - http://www.arm.linux.org.uk/
 maintainer of:  2.6 Serial core

Re: EXPORT_SYMBOL generates "is deprecated" noise

[Denis Vlasenko]

On Sunday 07 August 2005 22:15, Russell King wrote:
> On Sun, Aug 07, 2005 at 11:07:59AM -0700, Martin J. Bligh wrote:
> > I'm getting lots of errors like this nowadays:
> > 
> > drivers/serial/8250.c:2651: warning: `register_serial' is deprecated 
> > (declared at drivers/serial/8250.c:2607)
> > 
> > Which is just: "EXPORT_SYMBOL(register_serial);"
> > 
> > Sorry, but that's just compile-time noise, not anything useful.
> > Warning on real usages of it might be handy (we can go fix the users)
> > but not EXPORT_SYMBOL - we can't kill the export until the function
> > goes away. The more noise we have, the harder it is to see real errors 
> > and warnings.
> 
> I don't know why I bother with __deprecated - I haven't seen much
> evidence of the users of these functions being cleaned up, so I
> think we might as well just delete the functions and _force_ people
> to fix their code.  That unfortunately seems to be the only way to
> get things done in this day and age, which is rather sad if that's
> what it takes to kick people into action.

Was it ever different?
--
vda