Don't export __chown

Don't export __chown

[Andreas Schwab]

I think it is an oversight that __chown@GLIBC_2.1 is exported on i386.  It
is not explicitly mentioned in libc.map, and nobody has any business to
use such an internal name anyway.

Andreas.

2003-03-21  Andreas Schwab  <schwab@suse.de>

	* sysdeps/unix/sysv/linux/i386/chown.c: Don't define any __chown
	symbols.
	* abilist/libc.abilist: Remove __chown@GLIBC_2.1.

--- abilist/libc.abilist.~1.2.~	2003-03-20 10:38:29.000000000 +0100
+++ abilist/libc.abilist	2003-03-21 21:54:32.000000000 +0100
@@ -1477,7 +1477,6 @@ GLIBC_2.1.3 i.86-.*-linux.* powerpc-.*-l
  __cxa_finalize F
  __sigsuspend F
 GLIBC_2.1 i.86-.*-linux.*
- __chown F
  scalbln F
  scalblnf F
  scalblnl F
--- sysdeps/unix/sysv/linux/i386/chown.c.~1.17.~	2003-01-16 11:33:27.000000000 +0100
+++ sysdeps/unix/sysv/linux/i386/chown.c	2003-03-21 21:44:27.000000000 +0100
@@ -148,20 +148,13 @@ __chown_is_lchown (const char *file, uid
 #endif
 
 #if SHLIB_COMPAT (libc, GLIBC_2_0, GLIBC_2_1)
-strong_alias (__chown_is_lchown, _chown_is_lchown)
-compat_symbol (libc, __chown_is_lchown, __chown, GLIBC_2_0);
-compat_symbol (libc, _chown_is_lchown, chown, GLIBC_2_0);
+compat_symbol (libc, __chown_is_lchown, chown, GLIBC_2_0);
 #endif
 
 #ifdef __NR_lchown
-strong_alias (__real_chown, _real_chown)
-versioned_symbol (libc, __real_chown, __chown, GLIBC_2_1);
-versioned_symbol (libc, _real_chown, chown, GLIBC_2_1);
+versioned_symbol (libc, __real_chown, chown, GLIBC_2_1);
 libc_hidden_ver (__real_chown, __chown)
 #else
-strong_alias (__chown_is_lchown, __chown_is_lchown21)
-strong_alias (__chown_is_lchown, _chown_is_lchown21)
-versioned_symbol (libc, __chown_is_lchown21, __chown, GLIBC_2_1);
-versioned_symbol (libc, _chown_is_lchown21, chown, GLIBC_2_1);
+versioned_symbol (libc, __chown_is_lchown, chown, GLIBC_2_1);
 libc_hidden_ver (__chown_is_lchown, __chown)
 #endif

-- 
Andreas Schwab, SuSE Labs, schwab@suse.de
SuSE Linux AG, Deutschherrnstr. 15-19, D-90429 Nürnberg
Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."

Re: Don't export __chown

[Roland McGrath]

> I think it is an oversight that __chown@GLIBC_2.1 is exported on i386.  It
> is not explicitly mentioned in libc.map, and nobody has any business to
> use such an internal name anyway.

Agreed.  But we should wait to hear from others if anyone thinks this will
create a problem.

> 	* abilist/libc.abilist: Remove __chown@GLIBC_2.1.

Never hand-edit the .abilist files.  I won't merge such changes.  Always
use the update-abi/check-abi procedure, even though it just does the same
thing.  For i386, you don't need to bother because I will just regenerate
when I put the source changes in.  

Re: Don't export __chown

[Roland McGrath]

> I think it is an oversight that __chown@GLIBC_2.1 is exported on i386.

I agree.  This symbol was never in a Versions file.  Probably noone
realized that .symver definitions would result in exported symbols that
aren't listed in any Versions file (i.e. in the libc.map built from them)
and people only ever did visual checks on the Versions files.  Score
another one for check-abi.

I am inclined to change the generation of .map files to put "local: *;" at
the end of each set so that nothing not listed in a Versions file goes
through.