uapi/linux/a.out.h: don't use deprecated system-specific predefines.
PATCH 1/1 uapi/linux/a.out.h: don't use deprecated system-specific predefines.
uapi/linux/a.out.h uses a number of predefined macros that are deprecated because they're in the application namespace (e.g. '#ifdef linux' instead of '#ifdef __linux__'). This patch corrects all of them.
The primary reason this is worth bothering to fix, considering how obsolete a.out binary support is, is that the GCC build process considers this such a severe error that it will copy the header into a private directory and change the macro names, which causes future updates to the header to be masked. This header probably doesn't get updated very often anymore, but it is the _only_ uapi header that gets this treatment, so IMHO it is worth patching just to drive that number all the way to zero.
--- include/uapi/linux/a.out.h | 12 ++++++------ 1 file changed, 6 insertions(+), 6 deletions(-)
include/uapi/linux/a.out.h +6 -6
@@ -115,21 +115,21 @@ enum machine_type {/* Address of data segment in memory after it is loaded. Note that it is up to you to define SEGMENT_SIZE on machines not listed here. */ -#if defined(vax) || defined(hp300) || defined(pyr) +#if defined(__vax__) || defined(__hp300__) || defined(__pyr__)#define SEGMENT_SIZE page_size#endif -#ifdef sony +#ifdef __sony__#define SEGMENT_SIZE 0x2000#endif /* Sony. */ -#ifdef is68k +#ifdef __is68k__#define SEGMENT_SIZE 0x20000#endif -#if defined(m68k) && defined(PORTAR ) +#if defined(__m68k__) && defined(__PORTAR__)#define PAGE_SIZE 0x400#define SEGMENT_SIZE PAGE_SIZE#endif -#ifdef linux +#ifdef __linux__#ifndef __KERNEL__#include <unistd.h> #endif
@@ -260,7 +260,7 @@ struct relocation_info unsigned int r_extern:1;/* Four bits that aren't used, but when writing an object file it is desirable to clear them. */ -#ifdef NS32K +#ifdef __NS32K__ unsigned r_bsr:1; unsigned r_disp:1; unsigned r_pad:2;
Signature · 1 line
2.11.0
Quoted 10 lines
/* Address of data segment in memory after it is loaded.
Note that it is up to you to defineSEGMENT_SIZEon machines not listed here. */(quoted diff) +2
-2-#if defined(vax) || defined(hp300) || defined(pyr) +#if defined(__vax__) || defined(__hp300__) || defined(__pyr__) #define SEGMENT_SIZE page_size #endif -#ifdef sony +#ifdef __sony__ #define SEGMENT_SIZE 0x2000
Review
Quoted 10 lines
(quoted diff) +2
-2#endif /* Sony. */ -#ifdef is68k +#ifdef __is68k__ #define SEGMENT_SIZE 0x20000 #endif -#if defined(m68k) && defined(PORTAR) +#if defined(__m68k__) && defined(__PORTAR__) #define PAGE_SIZE 0x400 #define SEGMENT_SIZE PAGE_SIZE #endif
Can we please fix these conditionals so that only those relevant to Linux remain? e.g. something like the patch below:
include/uapi/linux/a.out.h +1 -25
@@ -112,24 +112,7 @@ enum machine_type {#define N_TXTADDR (x) (N_MAGIC(x) == QMAGIC ? PAGE_SIZE : 0)#endif -/* Address of data segment in memory after it is loaded. -Note that it is up to you to define SEGMENT_SIZE -on machines not listed here. */ -#if defined(vax) || defined(hp300) || defined(pyr) -#define SEGMENT_SIZE page_size -#endif -#ifdef sony -#define SEGMENT_SIZE 0x2000 -#endif /* Sony. */ -#ifdef is68k -#define SEGMENT_SIZE 0x20000 -#endif -#if defined(m68k) && defined(PORTAR ) -#define PAGE_SIZE 0x400 -#define SEGMENT_SIZE PAGE_SIZE -#endif - -#ifdef linux +/* Address of data segment in memory after it is loaded. */ #ifndef __KERNEL__#include <unistd.h> #endif
@@ -142,7 +125,6 @@ enum machine_type {#endif #endif #endif -#endif #define _N_SEGMENT_ROUND (x) ALIGN(x, SEGMENT_SIZE)
@@ -260,13 +242,7 @@ struct relocation_info unsigned int r_extern:1;/* Four bits that aren't used, but when writing an object file it is desirable to clear them. */ -#ifdef NS32K - unsigned r_bsr:1; - unsigned r_disp:1; - unsigned r_pad:2; -#else unsigned int r_pad:4; -#endif };#endif /* no N_RELOCATION_INFO_DECLARED. */
Quoted 3 lines · Christoph Hellwig
Can we please fix these conditionals so that only those relevant to Linux remain?
I was going for the minimal change since this is dusty-deck code that I can't practically test, but I have no objection to your more thorough approach. I don't see anything I would do differently.
zw
Quoted 8 lines · Zack Weinberg
Quoted 3 lines · Christoph Hellwig
Can we please fix these conditionals so that only those relevant to Linux remain?
I was going for the minimal change since this is dusty-deck code that I can't practically test, but I have no objection to your more thorough approach. I don't see anything I would do differently.
Yes, I think I'd just prefer to rip out the stale stuff, and remove the "#ifdef linux" entirely.
It's not like the kernel wants non-linux headers, and I don't see non-linux users using the kernel headers either.
Linus
Quoted 10 lines · Linus Torvalds
Quoted 8 lines · Zack Weinberg
Quoted 3 lines · Christoph Hellwig
Can we please fix these conditionals so that only those relevant to Linux remain?
I was going for the minimal change since this is dusty-deck code that I can't practically test, but I have no objection to your more thorough approach. I don't see anything I would do differently.
Quoted 5 lines
Yes, I think I'd just prefer to rip out the stale stuff, and remove the "#ifdef linux" entirely.
It's not like the kernel wants non-linux headers, and I don't see non-linux users using the kernel headers either.
Christoph's patch doesn't seem to have been applied yet. Is there anything I can do to help?
zw
Quoted 2 lines · Zack Weinberg
Christoph's patch doesn't seem to have been applied yet. Is there anything I can do to help?
A formal submission usually helps. I didn't want to post it under my name as it's mostly your work. How about this:
---
From: Zack Weinberg <[email protected]>
Subject: uapi/linux/a.out.h: don't use deprecated system-specific predefines.
uapi/linux/a.out.h uses a number of predefined macros that are deprecated because they're in the application namespace (e.g. '#ifdef linux' instead of '#ifdef __linux__'). This patch either corrects or just removes them if they are not applicable to Linux.
The primary reason this is worth bothering to fix, considering how obsolete a.out binary support is, is that the GCC build process considers this such a severe error that it will copy the header into a private directory and change the macro names, which causes future updates to the header to be masked. This header probably doesn't get updated very often anymore, but it is the _only_ uapi header that gets this treatment, so IMHO it is worth patching just to drive that number all the way to zero.
include/uapi/linux/a.out.h +1 -25
@@ -112,24 +112,7 @@ enum machine_type {#define N_TXTADDR (x) (N_MAGIC(x) == QMAGIC ? PAGE_SIZE : 0)#endif -/* Address of data segment in memory after it is loaded. -Note that it is up to you to define SEGMENT_SIZE -on machines not listed here. */ -#if defined(vax) || defined(hp300) || defined(pyr) -#define SEGMENT_SIZE page_size -#endif -#ifdef sony -#define SEGMENT_SIZE 0x2000 -#endif /* Sony. */ -#ifdef is68k -#define SEGMENT_SIZE 0x20000 -#endif -#if defined(m68k) && defined(PORTAR ) -#define PAGE_SIZE 0x400 -#define SEGMENT_SIZE PAGE_SIZE -#endif - -#ifdef linux +/* Address of data segment in memory after it is loaded. */ #ifndef __KERNEL__#include <unistd.h> #endif
@@ -142,7 +125,6 @@ enum machine_type {#endif #endif #endif -#endif #define _N_SEGMENT_ROUND (x) ALIGN(x, SEGMENT_SIZE)
@@ -260,13 +242,7 @@ struct relocation_info unsigned int r_extern:1;/* Four bits that aren't used, but when writing an object file it is desirable to clear them. */ -#ifdef NS32K - unsigned r_bsr:1; - unsigned r_disp:1; - unsigned r_pad:2; -#else unsigned int r_pad:4; -#endif };#endif /* no N_RELOCATION_INFO_DECLARED. */