Skip to contentSkip to conversation

Fix gcc 4 warnings on ia64

PATCH In review Andi Kleen 4 participants 6 messages Aug 16, 2005 – Aug 17, 2005

lore
Share & tools
lore

Andi Kleen

author
#0 lore raw

PATCH 1/1 Fix gcc 4 warnings on ia64

I was compiling an IA64 2.6.13rc6 kernel on a machine with gcc 4 to test something. The compile spew a incredibly large number of warnings. This patch fixes the most annoying ones from commonly included makefiles. In these two cases gcc was complaining that functions cannot be volatile or const.

There are a lot more warnings left, but they're in individual files. Most seem to be uninitialized variables because gcc 4 can track individual structure members for that now.

diff -u linux-2.6.13rc6/include/asm-ia64/mmu.h-o linux-2.6.13rc6/include/asm-ia64/mmu.h

include/asm-ia64/io.h renamed from include/asm-ia64/io.h-o +1 -1
diff -u linux-2.6.13rc6/include/asm-ia64/io.h-o linux-2.6.13rc6/include/asm-ia64/io.h
@@ -120,7 +120,7 @@
 	ia64_mfa();
 }
 
-static inline const unsigned long
+static inline unsigned long
 __ia64_get_io_port_base (void)
 {
 	extern unsigned long ia64_iobase;
include/asm-ia64/mmu_context.h renamed from include/asm-ia64/mmu_context.h-o +1 -1
diff -u linux-2.6.13rc6/include/asm-ia64/mmu_context.h-o linux-2.6.13rc6/include/asm-ia64/mmu_context.h
@@ -62,7 +62,7 @@
 	}
 }
 
-static inline mm_context_t
+static inline unsigned long
 get_mmu_context (struct mm_struct *mm)
 {
 	unsigned long flags;
#1 lore raw
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Andi" = Andi Kleen <[email protected]> writes:

Andi> I was compiling an IA64 2.6.13rc6 kernel on a machine with gcc 4 Andi> to test something. The compile spew a incredibly large number of Andi> warnings. This patch fixes the most annoying ones from commonly Andi> included makefiles. In these two cases gcc was complaining that Andi> functions cannot be volatile or const.

You actually need the const attribute on __ia64_get_io_port_base --- see the thread at
http://www.gelato.unsw.edu.au/archives/linux-ia64/0507/14658.html

--
Dr Peter Chubb http://www.gelato.unsw.edu.au peterc AT gelato.unsw.edu.au The technical we do immediately, the political takes *forever*

author
#2 lore raw
In reply to Andi Kleen
Quoted 11 lines · Peter Chubb
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line
Quoted 1 line

"Andi" = Andi Kleen <[email protected]> writes:

Andi> I was compiling an IA64 2.6.13rc6 kernel on a machine with gcc 4 Andi> to test something. The compile spew a incredibly large number of Andi> warnings. This patch fixes the most annoying ones from commonly Andi> included makefiles. In these two cases gcc was complaining that Andi> functions cannot be volatile or const.

You actually need the const attribute on __ia64_get_io_port_base --- see the thread at
http://www.gelato.unsw.edu.au/archives/linux-ia64/0507/14658.html

Ian is wrong here. const never did the same thing as __attribute__((const))

And for an inline __attribute__((const)) is a nop because the compiler always looks at the code inside the inline and decides based on that if the return value is constant or not.

-Andi

#3 lore raw
In reply to Andi Kleen
Quoted 2 lines

You actually need the const attribute on __ia64_get_io_port_base

cscope didn't show any users of this function ... so I thought perhaps the callers were in some macros that concatenated strings with ## to make the name, but I couldn't see any.

So I tried the big hammer and just deleted the function entirely to see what would complain, but nothing did (for compile with defconfig and tiger_defconfig).

So unless something very, very, very subtle is happening here, I'd say the fix would be to delete this function altogether. Rumours of better/worse code with/without the "const" modifier look to be unfounded.

Hint: we also store ia64_iobase in ar.k0 ... so perhaps this function has been obsolete for a while?

-Tony

author
#4 lore raw
In reply to Andi Kleen
Quoted 6 lines · Tony
Quoted 2 lines

You actually need the const attribute on __ia64_get_io_port_base

cscope didn't show any users of this function ... so I thought perhaps the callers were in some macros that concatenated strings with ## to make the name, but I couldn't see any.

You also need to fix the other function which had a bogus volatile.

BTW I bet the rationale in mmu.h for the volatile is bogus.

-Andi

#5 lore raw
In reply to Andi Kleen
Quoted 3 lines · Andi Kleen

And for an inline __attribute__((const)) is a nop because the compiler always looks at the code inside the inline and decides based on that if the return value is constant or not.

Not if the function contains an "asm", which is what used to be there.
It later got changed to reading a global, so I agree that the
__attribute_const__ is no longer necessary.

--david

Signature · 3 lines
Mosberger Consulting LLC, voice/fax: 510-744-9372,
http://www.mosberger-consulting.com/
35706 Runckel Lane, Fremont, CA 94536