remove wrong default for choice

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

support selecting which locales to build

Introduce UCLIBC_BUILD_MINIMAL_LOCALES and if selected
build only those locales.

Based on a patch by Bernhard Reutner-Fischer.

Signed-off-by: Marc Andre Tanner <mat@brain-dump.org>
Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

remember some TODOs for 0.9.31

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

disable _POSIX_SPAWN define

We don't provide spawn.h let alone any other spawn funcs/types, so don't
set up the _POSIX_SPAWN define that some packages (like vlc) check.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

build with -fmerge-all-constants

Glibc is already using this flag and it gives us a slight code shrink in
a few functions.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

test/plt: add a script to find PLT usage

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

libc: add hidden calls to pthread cleanup funcs

A lot of libc code calls the pthread cleanup funcs implicitly (for stdio)
which currently goes through the PLT.  Since we already have forwarding
symbols for these funcs, it's safe to declare the internal libc usage
hidden as a loaded libpthread will have the real symbols found.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

regex: call memcpy() ourselves

Call the hidden memcpy() ourselves otherwise gcc will emit a call to the
public memcpy() which goes through the PLT.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

sysctl: avoid inline initialization

Assign each field one by one rather than stack initialization as gcc will
call memset() to zero out the rest of the structure -- which we don't care
about as the field is unused and not seen outside of the libc.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

inet_ntop4: avoid inline initialization

We only need to set the first byte to 0, but gcc likes to zero out the
rest of the string with memset() when using this initialization style.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

sparc: use fputs to write to stderr

This also has the advantage of fputs() having a hidden alias while
puts does not.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

sparc: use HIDDEN_JUMPTARGET for errno

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

remove useless .gitignore

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

add .gitignore for test/ and extra/

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

test/dlopen: use pthread_once directly

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

linuxthreads: check TLS_DTV_AT_TP define correctly

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

mktemp does not depend on floats

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

gitignore: ignore install_dir/

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

Makerules: add a do_sed command

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

ldso: fixup missed variable rename ( tls_tpnt -> tpntp )

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

remove readelf from helptext

readelf was removed

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

malloc: handle size overflows in realloc()

The malloc() code checks the incoming size to make sure the header
adjustment doesn't cause overflow in the size storage.  Add the same
check to realloc() to catch stupid stuff like realloc(..., -1).

Reported-by: James Coleman <james.coleman@ubicom.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

sparc qops: unify & shrink

We don't really need to know the exact symbol that caused a failure as
long as we know where to start looking.  So unify the duplicate code
between all funcs and between the sparc variants.  This gives us a nice
code shrink of ~95%.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

ldso: drop duplicated/unused defines

We already include bits/fcntl.h for some of these defines, and most of
the bits/stat.h defines are unused.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

ldso: unify and cleanup _dl_mmap code

The interface should be straight forward now and the same for everyone.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

i386: use common INTERNAL_SYSCALL_ERROR_P

The common define has the same behavior, just formatted differently.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

ldso: define MAP_FAILED for everyone

This fixes build errors where common code has started using MAP_FAILED.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

ldso: simplify handling of extra _dl_lookup_hash() arg

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

ldso: clean up breakage in tls merge

The TLS merge 534661b91c9849 introduced multiple style problems as well as
random breakage:
- missing _dl_free
- incomplete parametrization of _dl_lookup_hash
- restore FDPIC handling in _dl_lookup_hash

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

ldso: fixup all the rest of the calls to _dl_find_hash

With TLS _dl_find_hash grew an extra param.
These archs don't have TLS reloc support yet, but they do need to
compile without it.

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

Fix SH bits/kernel_types.h for new kernel headers.

Recent SH kernel headers merge 32-bit and 64-bit headers, changing the
include guards on asm/posix_types.h in the process; update uClibc code
depending on those include guards.

Signed-off-by: Joseph Myers <joseph@codesourcery.com>

check stat("/etc/resolv.conf") for errors

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

fix cosmetic typo

s/dependancies/dependencies/g

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

libc/inet/resolv.c: reread resolv.conf if its mtime was changed

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

drop __BCC__ cruft from string code

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

check_ld some more flags

  gold does not currently implement these.

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

SUSV4: remove utime.h

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

add and use qstrip

  qstrip substitites quotes and strips the result.
  Avoiding single-quotes helps vim's poor syntax highlighting.

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

utime() is obsolescent in SUSv4

  LEGACY was removed for utimes() in SUSv4

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

ftw() is obsolescent in SUSv4

  Apps should switch to nftw()

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

_toupper, _tolower depend on SUSv4

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

link-time warning for obsolescent/removed network funcs

Signed-off-by: aldot <rep.dot.nop@gmail.com>

Fix building arm EABI

  The syscall() impl on aeabi comes from syscall-eabi.S thus we do not
  need the generic syscall() impl.

  Fixes:
    AS libc/sysdeps/linux/arm/syscall-eabi.os
    CC libc/sysdeps/linux/common/syscall.os
libc/libc_so.a(syscall.os): In function `syscall':
syscall.c:(.text+0x0): multiple definition of `syscall'
libc/libc_so.a(syscall-eabi.os):(.text+0x0): first defined here

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

resolv/crypt: convert to foo-y kbuild style

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

ldso: pull in missing bits/fcntl.h header for O_CLOEXEC

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

ldso: use O_CLOEXEC for cache file

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

clean up O_CLOEXEC handling

Drop the "#ifndef O_CLOEXEC" cruft, enable O_CLOEXEC in most fcntl.h
headers, and import __ASSUME_O_CLOEXEC from glibc.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

fcntl.h: drop unused/dead O_STREAMING

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

ldso: use MAP_FAILED with ld.so.cache

Rather than use our own custom casted -1 value, just use the MAP_FAILED
that common code already sets up for us for mmap().

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

fix coding style

  TODO: use likely/unlikely

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

commentary typo fix

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

trim whitespace

  and wrap superlong line.

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

whitespace cleanup

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

sparc: add stubs for more of the _Q* ops

Thanks to Jason Woodward <jason.woodward@timesys.com>

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

descriptive error message

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

mips: fix typo in elfinterp.c

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

Config.in: enable TLS config option only for nptl

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

errno.c: add tls versions of errno and h_errno

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

dl-support.c: add tls support

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

resolv.c: add support for per thread res_state

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

ldso/: tls support for dynamic linker

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

include/: add tls errno and res_state

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

mips: do not save variables in "hi" or "lo" across system calls

The kernel does not save these registers across system calls.  GCC 4.4
has gotten more aggressive about using them for temporary variables, so
this shows up as intermittent crashes if you use a recent compiler.

Signed-off-by: Daniel Jacobowitz <dan@codesourcery.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

sparc specific bits/socket.h

use socketcall for sparc32 since it does not have
the individual socket syscalls

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

doc/: typo fix

noted by Carlo Zinato <c.zinato@viscount.it>

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

add timerfd syscall and header

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>
Signed-off-by: Stephan Raue <mailinglists@openelec.tv>

sparc sigaction.c: fix the noted bugs by syncing up with glibc

Signed-off-by: Austin Foxley <austinf@cetoncorp.com>

sigpause: remove libc_hidden_proto/def

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

sigwait: remove __sigwait and __GI_sigwait symbols - they are unused

sigwait is not called from any uclibc function, so "hidden symbol"
trick is not needed on it. __sigwait also is never used,
and it's not clear why it even existed.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

raise: remove unused hidden weak symbol __raise

It is not used by anybody, otherwise users of dynamically linked uclibc
would notice - it is not exported from libuClibc-x.x.x.so

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

libc/misc/wctype/_wctype.c: make it more readable. no code changes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

remove a few more empty #if/#endif pairs

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

convert // comments to /**/; remove empty #if/#endif pairs. no code changes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

trim Experimentally off and uncommented hidden

  sed -i -e '/Experimentally off - /d' $(grep -rl "Experimentally off - " *)
  sed -i -e '/^\/\*[[:space:]]*libc_hidden_proto(/d' $(grep -rl "libc_hidden_proto" *)
  should be a nop

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

fix make {,install_}{,host}utils

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

document ARCH=

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

always define float_t as float

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

No need for a libresolv stub

  libresolv is just a stub anyway, default to not building it.

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

handle make ARCH= config

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

sys/mount.h: sync with latest glibc

Newer glibc adds a few more MNT_* flags, so import them.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

sh/user.h: clean up conflicts from kernel user.h

This patch makes sufficient modifications to the SH sys/user.h to be
able to build and run gdbserver with Linux kernel 2.6.30.

The problem is that both the kernel and library headers define the same
symbols, thus making it impossible in include both headers. The patch
undefines the kernel symbols in favour of the library symbols.

Signed-off-by: Andrew Stubbs <ams@codesourcery.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

daemon: fix up INTERNAL_SYSCALL() usage

Make sure we declare the error properly in case a port uses it, and fix the
invocation of exit().  Since clone() will be returning a pid, assume that
the value will not have the MSB set (negative) to simplify the error test a
bit more.  If gcc supports it, force this function to always be heavily
optimized in a bid to avoid stack usage as much as possible.

Signed-off-by: James Coleman <james.coleman@ubicom.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

setup attribute_optimize which comes with gcc-4.4+

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

Revert "default ?conf to native arch"

This reverts commit 6625518cd6894338937a74ca6b9187b7b8167b03.

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

sparc/sigaction.c: note possible bugs by comments. style fixes

no code is actually changed

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

strsignal.c: whitespace fixes, no code changes (verified with objdump)

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

time.c: style cleanup. no code changes (verified with objdump)

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

simpler and shorter read_TZ_file() helper

    text           data     bss     dec     hex filename
-   1109              8      76    1193     4a9 libc/misc/time/tzset.o
+   1095              8      76    1179     49b libc/misc/time/tzset.o

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

docs/defines.txt: add a note about possibly obsolete __LDBL_COMPAT

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

wordexp.c: cleanup - remove old hidden_proto's; remove unused parameter

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

wordexp.c: fix a bug where we might close stdout

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

fstatat: fix up behavior on 32/64 bit hosts

The fstatat() syscall is a little funky in that it sometimes changes name
between 32 and 64 bit hosts, but it should always operate on a 64bit stat
structure.  So for the fstatat() function, make sure we convert it from a
64bit kstat to a 32bit stat.

Along these lines, we need to restore the __xstat32_conv() function.

Reported-by: Timo Teräs <timo.teras@iki.fi>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

xstatconv: drop useless hidden code

The xstatconv.h header already sets up hidden prototypes for us, so no
need to duplicate the code.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

Revert "syscall.c: Use common syscall.c for ARM"

This reverts commit b1913a876059949e6c309bafade55e9425ef33fb.
OABI is still a requirement. So we override the common syscall
implementation specifically for arm port.

Signed-off-by: Khem Raj <raj.khem@gmail.com>

md5 passwords: reduce static usage

    text           data     bss     dec     hex filename
-   1875              0     120    1995     7cb libcrypt/md5.o
+   1855              0      35    1890     762 libcrypt/md5.o

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

rexec.c: style fixes, no code chages (verified with objdump)

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

rexec.c: small code shrink

    text           data     bss     dec     hex filename
-   1223              0       0    1223     4c7 libc/inet/rpc/rexec.o
+   1214              0       0    2239     8bf libc/inet/rpc/rexec.o

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

comment out "int rexecoptions"; make "char ahostbuf[]" static (was hidden)

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

do not save/restore errno around free() calls

In any non-buggy program free() does not fail.
And when it fails in a buggy program, the failure
is usually fatal (heap corruption and segfault).

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

do not pass 3rd param to open() which do not create files

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>