commit | 9fb22a3ec4ec5217fc29bb3ee3a0f1076098151f | [log] [tgz] |
---|---|---|
author | Elliott Hughes <enh@google.com> | Wed Oct 07 17:13:40 2015 -0700 |
committer | Elliott Hughes <enh@google.com> | Fri Oct 09 15:15:39 2015 -0700 |
tree | f5dabc7347b8aaa83675e4468edbccc657ae2fca | |
parent | cc56abbb007e090b830f510df6cb300ae12e1a89 [diff] |
Sync with upstream tzcode (2015g). This is quite a large patch because we haven't updated for some time, but the good news is that upstream is now thread-safe so a lot of our changes go away in this update and the remaining diff is a lot smaller. (Note that our whitespace still doesn't match upstream. I use diff -wub to compare. Upstream doesn't even really have a consistent style. New code seems to be two spaces, old code tabs.) From the intervening changelogs (eliding the changes that only affected the tools, which we don't use): 2014a: An uninitialized-storage bug in 'localtime' has been fixed. (Thanks to Logan Chien.) 2014b: 'zic' and 'localtime' no longer reject locations needing four transitions per year for the forseeable future. (Thanks to Andrew Main (Zefram).) 2014c: <None> 2014d: <None> 2014e: <None> 2014f: 'localtime', 'mktime', etc. now use much less stack space if ALL_STATE is defined. (Thanks to Elliott Hughes for reporting the problem.) Some lint has been removed when using GCC_DEBUG_FLAGS with GCC 4.9.0. 2014g: Unless NETBSD_INSPIRED is defined to 0, the tz library now supplies functions for creating and using objects that represent time zones. The new functions are tzalloc, tzfree, localtime_rz, mktime_z, and (if STD_INSPIRED is also defined) posix2time_z and time2posix_z. They are intended for performance: for example, localtime_rz (unlike localtime_r) is trivially thread-safe without locking. (Thanks to Christos Zoulas for proposing NetBSD-inspired functions, and to Alan Barrett and Jonathan Lennox for helping to debug the change.) If THREAD_SAFE is defined to 1, the tz library is now thread-safe. Although not needed for tz's own applications, which are single-threaded, this supports POSIX better if the tz library is used in multithreaded apps. Some crashes have been fixed when zdump or the tz library is given invalid or outlandish input. The tz library no longer mishandles leap seconds on platforms with unsigned time_t in time zones that lack ordinary transitions after 1970. The tz code now attempts to infer TM_GMTOFF and TM_ZONE if not already defined, to make it easier to configure on common platforms. Define NO_TM_GMTOFF and NO_TM_ZONE to suppress this. Unless the new macro UNINIT_TRAP is defined to 1, the tz code now assumes that reading uninitialized memory yields garbage values but does not cause other problems such as traps. If TM_GMTOFF is defined and UNINIT_TRAP is 0, mktime is now more likely to guess right for ambiguous time stamps near transitions where tm_isdst does not change. If HAVE_STRFTIME_L is defined to 1, the tz library now defines strftime_l for compatibility with recent versions of POSIX. Only the C locale is supported, though. HAVE_STRFTIME_L defaults to 1 on recent POSIX versions, and to 0 otherwise. tzselect -c now uses a hybrid distance measure that works better in Africa. (Thanks to Alan Barrett for noting the problem.) The C source code now ports to NetBSD when GCC_DEBUG_FLAGS is used, or when time_tz is defined. When HAVE_UTMPX_H is set the 'date' command now builds on systems whose <utmpx.h> file does not define WTMPX_FILE, and when setting the date it updates the wtmpx file if _PATH_WTMPX is defined. This affects GNU/Linux and similar systems. For easier maintenance later, some C code has been simplified, some lint has been removed, and the code has been tweaked so that plain 'make' is more likely to work. The C type 'bool' is now used for boolean values, instead of 'int'. The long-obsolete LOCALE_HOME code has been removed. The long-obsolete 'gtime' function has been removed. 2014h: The tz library's localtime and mktime functions now set tzname to a value appropriate for the requested time stamp, and zdump now uses this on platforms not defining TM_ZONE, fixing a 2014g regression. (Thanks to Tim Parenti for reporting the problem.) The tz library no longer sets tzname if localtime or mktime fails. An access to uninitalized data has been fixed. (Thanks to Jörg Richter for reporting the problem.) When THREAD_SAFE is defined, the code ports to the C11 memory model. A memory leak has been fixed if ALL_STATE and THREAD_SAFE are defined and two threads race to initialize data used by gmtime-like functions. (Thanks to Andy Heninger for reporting the problems.) 2014i: The time-related library functions now set errno on failure, and some crashes in the new tzalloc-related library functions have been fixed. (Thanks to Christos Zoulas for reporting most of these problems and for suggesting fixes.) If USG_COMPAT is defined and the requested time stamp is standard time, the tz library's localtime and mktime functions now set the extern variable timezone to a value appropriate for that time stamp; and similarly for ALTZONE, daylight saving time, and the altzone variable. This change is a companion to the tzname change in 2014h, and is designed to make timezone and altzone more compatible with tzname. The tz library's functions now set errno to EOVERFLOW if they fail because the result cannot be represented. ctime and ctime_r now return NULL and set errno when a time stamp is out of range, rather than having undefined behavior. Some bugs associated with the new 2014g functions have been fixed. This includes a bug that largely incapacitated the new functions time2posix_z and posix2time_z. (Thanks to Christos Zoulas.) It also includes some uses of uninitialized variables after tzalloc. The new code uses the standard type 'ssize_t', which the Makefile now gives porting advice about. 2014j: <None> 2015a: tzalloc now scrubs time zone abbreviations compatibly with the way that tzset always has, by replacing invalid bytes with '_' and by shortening too-long abbreviations. 2015b: Fix integer overflow bug in reference 'mktime' implementation. (Problem reported by Jörg Richter.) Allow -Dtime_tz=time_t compilations, and allow -Dtime_tz=... libraries to be used in the same executable as standard-library time_t functions. (Problems reported by Bradley White.) 2015c: <None> 2015d: <None> 2015e: <None> 2015f: <None> 2015g: localtime no longer mishandles America/Anchorage after 2037. (Thanks to Bradley White for reporting the bug.) On hosts with signed 32-bit time_t, localtime no longer mishandles Pacific/Fiji after 2038-01-16 14:00 UTC. The localtime module allows the variables 'timezone', 'daylight', and 'altzone' to be in common storage shared with other modules, and declares them in case the system <time.h> does not. (Problems reported by Kees Dekker.) On platforms with tm_zone, strftime.c now assumes it is not NULL. This simplifies the code and is consistent with zdump.c. (Problem reported by Christos Zoulas.) Change-Id: I9eb0a8323cb8bd9968fcfe612dc14f45aa3b59d2
The C library. Stuff like fopen(3)
and kill(2)
.
The math library. Traditionally Unix systems kept stuff like sin(3)
and cos(3)
in a separate library to save space in the days before shared libraries.
The dynamic linker interface library. This is actually just a bunch of stubs that the dynamic linker replaces with pointers to its own implementation at runtime. This is where stuff like dlopen(3)
lives.
The C++ ABI support functions. The C++ compiler doesn't know how to implement thread-safe static initialization and the like, so it just calls functions that are supplied by the system. Stuff like __cxa_guard_acquire
and __cxa_pure_virtual
live here.
The dynamic linker. When you run a dynamically-linked executable, its ELF file has a DT_INTERP
entry that says "use the following program to start me". On Android, that's either linker
or linker64
(depending on whether it's a 32-bit or 64-bit executable). It's responsible for loading the ELF executable into memory and resolving references to symbols (so that when your code tries to jump to fopen(3)
, say, it lands in the right place).
The tests/
directory contains unit tests. Roughly arranged as one file per publicly-exported header file.
The benchmarks/
directory contains benchmarks.
Adding a system call usually involves:
As mentioned above, this is currently a two-step process:
This is fully automated:
If you make a change that is likely to have a wide effect on the tree (such as a libc header change), you should run make checkbuild
. A regular make
will not build the entire tree; just the minimum number of projects that are required for the device. Tests, additional developer tools, and various other modules will not be built. Note that make checkbuild
will not be complete either, as make tests
covers a few additional modules, but generally speaking make checkbuild
is enough.
The tests are all built from the tests/ directory.
$ mma $ adb remount $ adb sync $ adb shell /data/nativetest/bionic-unit-tests/bionic-unit-tests32 $ adb shell \ /data/nativetest/bionic-unit-tests-static/bionic-unit-tests-static32 # Only for 64-bit targets $ adb shell /data/nativetest64/bionic-unit-tests/bionic-unit-tests64 $ adb shell \ /data/nativetest64/bionic-unit-tests-static/bionic-unit-tests-static64
The host tests require that you have lunch
ed either an x86 or x86_64 target.
$ mma $ mm bionic-unit-tests-run-on-host32 $ mm bionic-unit-tests-run-on-host64 # For 64-bit *targets* only.
As a way to check that our tests do in fact test the correct behavior (and not just the behavior we think is correct), it is possible to run the tests against the host's glibc. The executables are already in your path.
$ mma $ bionic-unit-tests-glibc32 $ bionic-unit-tests-glibc64
For either host or target coverage, you must first:
$ export NATIVE_COVERAGE=true
bionic_coverage=true
in libc/Android.mk
and libm/Android.mk
.$ mma $ adb sync $ adb shell \ GCOV_PREFIX=/data/local/tmp/gcov \ GCOV_PREFIX_STRIP=`echo $ANDROID_BUILD_TOP | grep -o / | wc -l` \ /data/nativetest/bionic-unit-tests/bionic-unit-tests32 $ acov
acov
will pull all coverage information from the device, push it to the right directories, run lcov
, and open the coverage report in your browser.
First, build and run the host tests as usual (see above).
$ croot $ lcov -c -d $ANDROID_PRODUCT_OUT -o coverage.info $ genhtml -o covreport coverage.info # or lcov --list coverage.info
The coverage report is now available at covreport/index.html
.
Bionic's test runner will run each test in its own process by default to prevent tests failures from impacting other tests. This also has the added benefit of running them in parallel, so they are much faster.
However, this also makes it difficult to run the tests under GDB. To prevent each test from being forked, run the tests with the flag --no-isolate
.
This probably belongs in the NDK documentation rather than here, but these are the known ABI bugs in LP32:
time_t
is 32-bit. http://b/5819737
off_t
is 32-bit. There is off64_t
, but no _FILE_OFFSET_BITS
support. Many of the off64_t
functions are missing in older releases, and stdio uses 32-bit offsets, so there's no way to fully implement _FILE_OFFSET_BITS
.
sigset_t
is too small on ARM and x86 (but correct on MIPS), so support for real-time signals is broken. http://b/5828899