We've never really used __restrict: only <string.h> and <stdio.h> (which
are still very similar to upstream BSD headers) consistently have these
annotations. Neither clang nor GCC warns for trivial cases, and there's
little obvious documentation benefit.
Bug: http://b/30833514
Test: builds
Change-Id: I3e4384281865475d0c55d764b546d8166419ee31
I plan on having one review per file for the enable_if->diagnose_if
FORTIFY migration. Having this in means that no one review is dependent
on another.
Bug: 12231437
Test: m.
Change-Id: Ic0b07d7f7f6782e371c8792eb1e40cdfa32d3e35
At least one warning only triggers in files that are included, instead
of being passed directly. Switch to compiling with -include, and fix
the resulting warnings.
Bug: https://github.com/android-ndk/ndk/issues/474
Test: mma -j && versioner
Test: python tools/versioner/run_tests.py
Change-Id: I784698c18540c9cc30f372f279a1cec1d75721ea
This patch cleans up our standard headers by moving most of the FORTIFY
cruft out in to its own sandbox. In order to include the *_chk and
*_real declarations, you can either enable FORTIFY, or `#define
__BIONIC_DECLARE_FORTIFY_HELPERS`.
Both sys/select.h and strings.h are explicitly ignored by this patch.
Both of these files have very small __BIONIC_FORTIFY blocks, and don't
define any actual FORTIFY'ed functions (just macros, and 3 *_chk
functions).
This patch also makes the versioner ignore the FORTIFY implementation
headers, since we're guaranteed to pick the FORTIFY'ed headers up when
looking at the regular headers. (...Not to mention that making the
FORTIFY'ed headers freestanding would be annoying to do and maintain for
~no benefit).
We bake the knowledge of where FORTIFY headers live directly into the
versioner. We could go with a more general approach (e.g. adding an -X
IGNORED_FILE flag that tells the versioner to ignore
$HEADER_PATH/$IGNORED_FILE), but we'd then have to repeat that for every
test, every manual invocation of the versioner, etc. for no benefit
that's obvious to me.
Bug: 12231437
Test: m checkbuild on bullhead internal master + CtsBionicTestCases. no
new errors.
Change-Id: Iffc0cc609009b33d989cdaddde0a809282131a5b
Old versions of Android called these fdprintf and vfdprintf out of
fears that the glibc names would collide with user debug printfs.
Allow users to just use dprintf and vfdprintf on any version by
renaming those calls to their legacy equivalents if needed.
Test: built trivial NDK module targeting android-14 and using dprintf
Test: make checkbuild
Bug: https://github.com/android-ndk/ndk/issues/72
Change-Id: I90de149278f931380418536abaef47c5cee5c195
I've added some things like __ANDROID_MAJOR__ to an ndk-version.h, but
that is only in the NDK itself and so doesn't help the platform. Add
__ANDROID_NDK__ to identify that you're building for the NDK and not
the platform.
Test: make checkbuild
Bug: https://github.com/android-ndk/ndk/issues/407
Change-Id: I2d1f1c28e3764e4e658cf675b290b7a17253ee33
It's not usable for cases where the off_t and off64_t functions became
available in different API levels, so it's not as big an improvement as
we'd hoped, but it cleans up several headers and should be usable for
any future additions (though recent Linux additions have _only_ supported
off64_t anyway).
Bug: N/A
Test: builds
Change-Id: I18b00f30666079d9d12d5b9c0ed916076d6c3641
Convenience macro for __RENAME if __USE_FILE_OFFSET64 mode. Lets us
avoid duplicating all the doxygen comments in frameworks headers.
Test: make checkbuild
Bug: https://github.com/android-ndk/ndk/issues/459
Change-Id: Ica44f22b2f1596e484694006c0926d94d16187b5
We had several bugs filed saying "if I set _FILE_OFFSET_BITS=64 when
targeting an API < L, various functions are missing". Instead of
saying "yes, they are", we quietly just modified the header files to
expose the non-64-bit variants. This makes no sense. We can't just say
"oh, yeah, we don't have a version of this function that agrees with
your calling code about how large off_t is, but here's a version that
doesn't: I'm sure it'll be fine".
_FILE_OFFSET_BITS=64 on Android LP32 has always been a game of chance,
but that game should be "are all the functions my code needs available
at compile time?", not "will my code actually work at run time?".
Bug: https://github.com/android-ndk/ndk/issues/449
Bug: https://github.com/android-ndk/ndk/issues/442
Bug: https://github.com/android-ndk/ndk/issues/333
Bug: https://github.com/android-ndk/ndk/issues/332
Bug: https://github.com/android-ndk/ndk/issues/324
Test: builds
Change-Id: Ib095251d3e21e77ed50cc3575388107fecec4ecd
x86 needs the <sys/user.h> additions, x86-64 the <sys/procfs.h> ones.
Bug: N/A
Test: ndkports x86 and x86_64 binutils-2.28
Change-Id: I2e4738d3f8343d12db0536ca77045e2a0ccc6596
* TTYDEF_OFLAG referred to the BSD OXTABS constant instead of the Linux XTABS,
so references to it wouldn't compile.
* CEOL was incorrect (compared to `stty -a`).
* ttydefchars wouldn't compile (removed rather than attempt to fix something
of unclear utility).
Bug: http://b/63145226
Test: new test
Change-Id: I823e5eaab789cafd2eeea03f77ea553346822f74
We pretty much do these days. And apparently some GNU tools turn on obsolete
backwards-compatible behaviors otherwise.
Bug: http://b/63145226
Test: builds
Change-Id: I9a8af860768cc2c10548f172d1e8b4fba3e5115b
We've had <arpa/telnet.h> for years, despite being slightly more useless
than <arpa/ftp.h> and <arpa/tftp.h>. The inetutils package expects these
to be available.
Bug: http://b/63145226
Test: builds
Change-Id: I1e3c7421779e965b3342d681fefb8dc2561b604d
As in other places, only expose the constants at API levels where
we're also exposing the functions (to cope with "poor man's configure"
where an author assumes the presence of the #define to imply the
availability of the function).
Bug: http://b/63145226
Test: builds
Change-Id: I524f7288513aa6c35479c550e748a619397929b0
Before this change we have the old NDK inline termios functions with the
modern constants. Unfortunately the old NDK inline functions relied on
hacking the constants. Fix things by sharing the implementation between
the platform and the NDK headers.
Bug: https://github.com/android-ndk/ndk/issues/441
Test: ran tests
Change-Id: I2773634059530bc954167f29c4783413a2294d5a
<strings.h>'s ffs used to work by accident. In the past, <strings.h> used
to incorrectly declare ffs for all platforms at all API levels. In the
unified headers, there's no such declaration for x86 before API 18 (which
makes sense, because that function was missing on x86 until then).
But as long as there was a declaration for ffs, the compiler just inlined
__builtin_ffs. There was no problem at link time because the compiler didn't
actually add a reference to the missing ffs symbol.
Restore the old behavior by manually instructing the compiler to inline its
builtin in these cases.
Bug: https://github.com/android-ndk/ndk/issues/439
Test: built new NDK 'ffs' test
Change-Id: I840e99f237c86f7cb028a0f67aaa8c6ff3eda245
Rather than do this in libandroid_support, we may as well just stick it with
the other historical stdlib workarounds in bionic itself...
Bug: N/A
Test: built new NDK test
Change-Id: Ia5cf4010581eb79d4adf924e87d0bc050b9e2839
These still won't get guards added by the preprocessor, because it
compiles with C-only.
Bug: https://github.com/android-ndk/ndk/issues/440
Test: treehugger
Change-Id: I893b345e528ed1b761e0db00700037411bbb8b78
Building ruby actually trips over both of these:
* if the RTLD_ constants aren't #defined, it uses its own incorrect values.
* if the REG_ constants aren't #defined, it confuses x86 with x86-64.
In all other places where we have enums in our headers, we already match
existing glibc practice.
Bug: http://b/62531921
Test: builds
Change-Id: I5b3aab25a1a24611bdc58f2eda4104a78e9f841c
No other C library expose these, and I couldn't find any callers.
Bug: http://b/62531921
Test: builds
Change-Id: I4a3505bc0897286a4036c48066b98d16665b573a
In a similar style to some of our other "not really, but enough" headers
like <sys/vt.h>.
Bug: N/A
Test: build GNU dd or BSD dd with a standalone toolchain
Change-Id: I8fbd1aac1d97e24b05e7aae8a55666300b5bf1ed
This breaks code that declares `errno` itself for whatever crazy reason:
b.c:22:12: error: conflicting types for '__errno'
extern int errno;
^
/usr/local/google/ndkports/toolchain/bin/../sysroot/usr/include/errno.h:47:20: note: expanded from macro 'errno'
#define errno (*__errno())
^
/usr/local/google/ndkports/toolchain/bin/../sysroot/usr/include/errno.h:44:15: note: previous declaration is here
volatile int* __errno(void) __attribute_const__;
^
Bug: N/A
Test: built various bits of GNU source
Change-Id: I27c03bf3bde419a001f98f1ea6c267c847f31271
This causes GNU make to fail to build:
main.o:main.c:(.text+0x18c): more undefined references to `bsd_signal' follow
Bug: N/A
Test: built GNU make with a hacked standalone toolchain
Change-Id: Icdfb4cf674df8abed44f6671454bed9f8d68e74b
The proper API for this isn't available until L, so expose this for
API levels earlier than that.
Test: make checkbuild
Bug: https://github.com/android-ndk/ndk/issues/420
Change-Id: I382b8f557be9530f3e13aaae353b4a6e7f9301ab
boost (1.64.0 at least) assumes that if you have POSIX_MADV_NORMAL, you have
posix_madvise. With unified headers, this isn't true. Rather than make life
harder for projects that don't use configure, just make it so. We already
applied similar workarounds for epoll_create1 and inotify_init1.
Bug: https://github.com/android-ndk/ndk/issues/395
Test: built boost (long story!)
Change-Id: I5d2d8de7b30921dde913251d35dcd249a2876f94
<sys/types.h> unconditionally includes <sys/sysmacros.h>.
<sys/sysmacros.h> defines major, minor, and makedev. In the deprecated NDK
headers, these were inline functions. In the unified headers, they are
function-like macros.
The inline functions would only collide with another function called major,
minor, or makedev but the macros will replace anything with a function-like
form such as definitions of class methods called major, minor, or makedev.
This causes code which has such definitions to fail to compile when moving
from the deprecated headers to the unified headers.
Resolve this by removing the transitive dependency. Folks who want
<sys/sysmacros.h> can ask for it explicitly (it's been in both bionic and
glibc for years, though I don't think macOS has it [as usual]).
Bug: https://github.com/android-ndk/ndk/issues/398
Test: builds
Change-Id: If2921c7b5d979c3066f199f22c64d4d2f7bf6632
Some third-party code uses the existence of IN_CLOEXEC/IN_NONBLOCK to
detect the availability of inotify_init1. This is not correct, since
`syscall(__NR_inotify_init1, IN_CLOEXEC)` is still valid even if the C
library doesn't have that function, but for the time being we don't
want to harm adoption to the unified headers. We'll avoid defining
IN_CLOEXEC and IN_NONBLOCK if we don't have inotify_init1 for the time
being, and maybe revisit this later.
Test: make checkbuild
Bug: https://github.com/android-ndk/ndk/issues/394
Change-Id: Iefdc1662b21045de886c7ad1cbeba6241163d943
Some third-party code uses the existence of EPOLL_CLOEXEC to detect
the availability of epoll_create1. This is not correct, since having
up-to-date UAPI headers says nothing about the C library, but for the
time being we don't want to harm adoption to the unified headers.
We'll undef EPOLL_CLOEXEC if we don't have epoll_create1 for the time
being, and maybe revisit this later.
Test: make checkbuild
Bug: https://github.com/android-ndk/ndk/issues/302
Bug: https://github.com/android-ndk/ndk/issues/394
Change-Id: I8ee9ce62768fb174070ec51d114f477389befc4a
We've had pthread_*_setpshared for long enough that there are no
__INTRODUCED_IN guards.
Found because fio's configure script was confused by this.
Bug: N/A
Test: ran tests
Change-Id: I07b4d937741e4dcd7e615f2371b17c827341917a
This is a bit bogus because it's been removed from glibc (though not
thoroughly) and is never useful on Android (because the system calls
in question are compiled out of Android kernels, and SELinux would
disallow them even if you weren't running an Android kernel). This
also means that on glibc you need to include <linux/sem.h> for this
and on bionic you need <sys/sem.h> (and for either if you #include
the other file, you won't get this union).
Bug: https://github.com/android-ndk/ndk/issues/400
Test: added new test
Change-Id: I47f721da77515531f616d6ad8479bfbc9b60ee47
Add the mallopt function, and only a single option so far.
Bug: 36401135
Test: Built and booted bullhead.
Test: Ran jemalloc unit tests.
Test: Ran bionic unit tests.
Test: Ran a test that allocated and free'd a large piece of memory,
Test: and verified that after changing the parameter, the PSS
Test: sticks around (decay timer set to 1), the PSS is purged (decay
Test: timer set to 0).
Change-Id: I6927929b0c539c1023d34772d9e26bb6a8a45877
Mistakenly removed years ago, along with its genuinely duplicated companions.
This patch also redefines one of those rather than use the Linux uapi header
definition, which leads to signed/unsigned warnings in calling code.
Bug: https://github.com/android-ndk/ndk/issues/377
Test: ran tests
Change-Id: Ib25b17949f04969f12b945ca88ab87d080677cf9
After the new rebase of Clang, we now supports alloc_size attribute, so
we can remove the warning about switching to alloc_size once Clang has
support.
Compiler.setInvocation has changed argument type from raw pointer to
shared pointer. Add version check here so that we can build under either
old rebase or the new rebase.
Test: build
Bug: 37423073
Change-Id: I4563eaf93bae6c59a4a19318f8caa92bd361b3ab
BSD thinks you should only get the relocation constants for your target
architecture, but it's often useful to have them all available at once.
Rearrange the headers to enable that.
Also update the (modified) NetBSD files to CVS HEAD.
Also remove the unused BSDism R_TYPE.
Bug: N/A
Test: builds
Change-Id: Iad5ef29192a732696e2b36af35144a9ca116aa46
Apparently __strncpy_real redirects to strcpy instead of strncpy for GCC
builds. This is bad, and it confused a project compiled by GCC that uses
strcpy.
I audited all of the FORTIFY-related __RENAMEs both manually and with a
script I hacked together; they all look correct aside from this one.
FWIW, with the cleaned up clang FORTIFY, lots of the `void foo()
__RENAME(foo)` will just become `void foo()` if this last thing I'm
working on gets through review.
Bug: 35329524
Test: Previously broken project now seems to work. Bullhead and Ryu
both build + boot.
Change-Id: Ib18ee3535ae31eb7e8ae846dc012f9b64cac80bf
AP_FPUCW is referenced by valgrind.
Bug: https://bugs.kde.org/show_bug.cgi?id=339945
Test: valgrind builds with one less hack
Change-Id: I7146f3709286662108cc82ba3c55ce935375a1f1
Netflix was using this, and looking the header file, although
__system_property_find_nth has been available since the beginning of time,
__system_property_foreach only appeared in 16. So anyone who wants to run
on pre-JellyBean devices would want to use __system_property_find_nth.
It's pretty much a one-liner in terms of __system_property_foreach anyway,
so it doesn't cost us anything to keep it.
Also restore slightly better tests than we originally removed.
Bug: http://b/36566667
Test: ran tests
Change-Id: Id268c2c2e848da17bb0a5a5420af234d9dcb829a
This gives more useful diagnostics if clang needs to mention these
functions (or a parameter of them).
Bug: 36984245
Test: m on bullhead completes successfully.
Change-Id: I17c2b624d08bc9dd3f08185b30029ed0c49ebb08
libcxx provides const-correct overloads for a few string.h functions.
These overloads use clang's enable_if attribute, so they're preferred
over our FORTIFY'ed equivalents.
This weakens _FORTIFY_SOURCE=2 when used with some of these functions,
since clang needs to see __pass_object_size in order to pass an accurate
result for __builtin_object_size(s, 1) at a callsite. Since those
functions don't have __pass_object_size on their params, clang can't do
that. This makes LLVM lower the __builtin_object_size calls, which means
we get the same result as __builtin_object_size(s, 0).
We have to provide all of the overloads in Bionic, since enable_if is
only used to disambiguate overloads with (otherwise) the same type. In
other words:
// overload 1
char *strchr(const char *, int s) __attribute__((enable_if(1, "")));
// overload 2
char *strchr(char *, int s);
void foo() {
char cs[1] = {};
strchr(static_cast<const char *>(cs), '\0'); // calls overload #1.
strchr(cs, '\0'); // calls overload #2.
}
Bug: 34747525
Test: m checkbuild on bullhead internal master + AOSP. vts -m
BionicUnitTests passes on both. Surprisingly, the only code that this
seems to break is contained in Bionic.
Change-Id: Ie406f42fb3d1c5bf940dc857889876fc39b57c90
I dunno why I used __bos0 in the first place; clang's strrchr (and
GCC's strchr+strrchr) both use __bos.
Bug: 34747525
Test: m. Device still boots. cts -m BionicUnitTests shows no new
failures.
Change-Id: Ifec0e05a6a1144fa3e3ac70562af3ec57c09c194
While this change is correct, GNU libstdc++ 4.9 isn't able to handle a
standard compliant C library. Its <cmath> will `#undef isnan` from
math.h and only adds the function overloads to the std namespace,
making it impossible to use both <cmath> (which gets included by a
lot of other standard headers) and ::isnan.
We're going to have to revert this until we can start turning down
support for gnustl.
This reverts commit e76ee993ff.
Bug: https://code.google.com/p/android/issues/detail?id=271629
Test: make checkbuild
Change-Id: I394f50271430e78ab801d85c3ee4e87019eda6af
