summary refs log tree commit diff
path: root/sysdeps/unix/sysv/linux/kernel-features.h
diff options
context:
space:
mode:
authorJoseph Myers <joseph@codesourcery.com>2018-12-18 13:35:39 +0000
committerJoseph Myers <joseph@codesourcery.com>2018-12-18 13:35:39 +0000
commit646ce7e0be9218f644ab50681b4d5a13d1050dd4 (patch)
tree846db25ccb66604cb43058279676a0bee414b16c /sysdeps/unix/sysv/linux/kernel-features.h
parent80190d2b0e3f48d973724218f37d2da5bf1a20ab (diff)
downloadglibc-646ce7e0be9218f644ab50681b4d5a13d1050dd4.tar.gz
glibc-646ce7e0be9218f644ab50681b4d5a13d1050dd4.tar.xz
glibc-646ce7e0be9218f644ab50681b4d5a13d1050dd4.zip
Remove __ASSUME_ST_INO_64_BIT.
kernel-features.h has a macro __ASSUME_ST_INO_64_BIT, with a comment
"However, SH is lame, and still does not have a 64-bit inode field.".

The macro is, in fact, defined to 0 by Alpha as well as SH.  The Alpha
case is, however, trivially useless: none of the files that test
__ASSUME_ST_INO_64_BIT are built for Alpha (which gained kernel
support for stat64 syscalls, with a 64-bit st_ino field, in Linux
2.6.4; the define to 0 for Alpha in glibc predates that).

The SH kernel gained support for a 64-bit st_ino in struct stat64 in
commit 760bcb1deec13c50e20399c84cb6a8ea41cc2820 ("sh: Fix fstatat64()
syscall."), which is in Linux 2.6.22 and later.  So the redefinition
of __ASSUME_ST_INO_64_BIT to 0 is of no use for SH either; three of
the files testing it do so immediately after a stat64-family syscall
has been used, which will always have set the 64-bit st_ino correctly
(in addition to the 32-bit __st_ino), while the relevant code
__xstat32_conv executes only after such a syscall in the function
calling __xstat32_conv.

Thus this patch removes __ASSUME_ST_INO_64_BIT and code testing it.
Removing the useless [!__ASSUME_ST_INO_64_BIT] code in __xstat32_conv
renders the [_HAVE_STAT64___ST_INO] and [!_HAVE_STAT64___ST_INO] cases
around it identical, so that conditional is also removed.

Tested compilation with build-many-glibcs.py for its Alpha and SH
configurations; also ran the glibc testsuite for x86_64 and x86.

	* sysdeps/unix/sysv/linux/kernel-features.h
	(__ASSUME_ST_INO_64_BIT): Remove macro definition.
	* sysdeps/unix/sysv/linux/alpha/kernel-features.h
	(__ASSUME_ST_INO_64_BIT): Do not undefine and define.
	* sysdeps/unix/sysv/linux/sh/kernel-features.h
	(__ASSUME_ST_INO_64_BIT): Likewise.
	* sysdeps/unix/sysv/linux/fxstat64.c: Do not include
	<kernel-features.h>.
	(___fxstat64) [_HAVE_STAT64___ST_INO && !__ASSUME_ST_INO_64_BIT]:
	Remove conditional code.
	* sysdeps/unix/sysv/linux/lxstat64.c: Do not include
	<kernel-features.h>.
	(___lxstat64) [_HAVE_STAT64___ST_INO && !__ASSUME_ST_INO_64_BIT]:
	Remove conditional code.
	* sysdeps/unix/sysv/linux/xstat64.c: Do not include
	<kernel-features.h>.
	(___xstat64) [_HAVE_STAT64___ST_INO && !__ASSUME_ST_INO_64_BIT]:
	Remove conditional code.
	* sysdeps/unix/sysv/linux/xstatconv.c: Do not include
	<kernel-features.h>.
	(__xstat32_conv) [_HAVE_STAT64___ST_INO]: Remove conditional code.
	[!_HAVE_STAT64___ST_INO]: Make code unconditional.
Diffstat (limited to 'sysdeps/unix/sysv/linux/kernel-features.h')
-rw-r--r--sysdeps/unix/sysv/linux/kernel-features.h4
1 files changed, 0 insertions, 4 deletions
diff --git a/sysdeps/unix/sysv/linux/kernel-features.h b/sysdeps/unix/sysv/linux/kernel-features.h
index 0ac1616b33..911eb96fbe 100644
--- a/sysdeps/unix/sysv/linux/kernel-features.h
+++ b/sysdeps/unix/sysv/linux/kernel-features.h
@@ -37,10 +37,6 @@
    introduced.  If somebody cares these values can afterwards be
    corrected.  */
 
-/* The changed st_ino field appeared in 2.4.0-test6.  However, SH is lame,
-   and still does not have a 64-bit inode field.  */
-#define __ASSUME_ST_INO_64_BIT		1
-
 /* The statfs64 syscalls are available in 2.5.74 (but not for alpha).  */
 #define __ASSUME_STATFS64	1