| Age | Commit message (Collapse) | Author |
|---|
|
Elliott said:
Works around a clang bug in the version of clang shipping with Android T
that causes a miscompile in xstdio_create() on x86-64, that oddly only
affects __attribute__((__noreturn__)) and not _Noreturn!
Bug: http://b/225806996
Test: manual inspection of generated code on x86-64
Change-Id: I12302d2300fc454f92352b43090745c04ed66fd0
|
|
|
|
Android's host tools currently build with glibc from 2012, but even if
we were more up to date (or had already switched to musl for host tools,
which hopefully will be the case by this time next year), it was added
to glibc recently enough to miss the "seven year window".
|
|
so "command -123" isn't an unknown option error.
|
|
Also in lib/ add bufget name variants for password and group.
|
|
Teach chomp() to remove any number of trailing \r\n
|
|
|
|
|
|
|
|
(The kernel's been doing this since 2019, but older kernels may not, so...)
|
|
Fallback to read+write when copy_file_range failed, as both fd_in and fd_out must refer to regular files.
Signed-off-by: LongPing Wei <weilongping@huawei.com>
|
|
|
|
|
|
seperately. (Nothing else in lib/ has to rebuild when command list changes.)
|
|
|
|
|
|
|
|
interrupted long enough ago it would be easier to restart from scratch.
It was never used and hasn't been touched in years.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Also add QUIET macro to shut up gcc's "isn't actually used uninitalized"
warnings, instead of doing the "int x=x;" trick.
|
|
non-threaded programs, and neither bionic nor musl have it. (This means glibc always does the HASTIMERS syscall wrap.)
|
|
|
|
Existing directories should be ignored without errors only with -p flag (according
to POSIX).
Signed-off-by: Mikhail Kashkarov <m.kashkarov@samsung.com>
|
|
Also modernize pidof.c, even though the bug was in lib.
|
|
|
|
For some reason "configure" is explicitly NOT passing
LDOPTIMIZE="-Wl,--gc-sections" to the mac build. I don't know why,
but it means the linker includes unused functions in the binary,
and thus complains about unused dependencies being missing.
|
|
|
|
(Need to test this on FreeBSD too...)
|
|
|
|
Note that these aren't equivalent to the CLOCK_MONOTONIC timer we
have on Linux, but short of spinning up another thread ourselves I
can't see how to implement that on macOS at present.
|
|
On a desktop or server the difference between the realtime and monotonic
clocks doesn't matter much, but we've seen cases on Android where
something's run under timeout(1) but killed before it's had a chance to
do anything because the device went into sleep shortly afterwards and by
the time it came back, the timeout had elapsed (as far as the realtime
clock is concerned) but the process had only had a tiny fraction of the
timeout where the system wasn't suspended.
This patch also adds an xparsetimespec() function; at this point the
only remaining users of xparsetime() are in lib, so maybe that should
become static, at least until we have another need for it?
A bigger question is whether timeout(1) needs to offer the user a choice
between the two different clocks? Although monotonic is usally the right
default choice on Android, and I don't have a specific counter-example
to hand, I can imagine that someone might actually mean "wall clock time"
rather than "cpu clock time" when setting a timeout...
Luckily, timer_create() lets us trivially choose between both clocks
if so.
|
|
Obviously we're never going to be able to build strace for macOS,
but right now the reference to the struct was breaking even a macOS
build with strace disabled! From poking at building for arm on the
Raspberry Pi over the weekend, changes like this will be needed
there too (because different architectures have different _names_
for this struct, not just different contents).
Tested on macOS (where it juts fixes the build) and on x86-64 (where
strace continues to function as before).
|
|
This is most notably lacking -p and -f, and only covers a small subset
of system calls, but it's "useful" in the sense that you can strace
something like toybox date, say, and see everything.
There's a fundamental assumption here that we don't need to worry about
multiple personalities. (This is what makes "real" strace's build so
complex, and why Android has _never_ actually shipped a fully correct
strace build.) Given that the desktop's already been 64-bit only
for a while, microcontrollers are likely to stay 32-bit only for a
while, and the odd one out -- mobile -- is moving to be 64-bit only
(https://www.arm.com/blogs/blueprint/64-bit), so by the time this is
ready to graduate from pending, I'm assuming no-one will care!
|
|
|
|
|
|
|
|
Allow num_to_sig(0) so kill doesn't disallow -0 because it can't
round-trip, and then teach list_signals() to avoid the special case of
signal 0 rather than assume that all the signals it shouldn't print
don't have names now signal 0 has the name "0".
|
|
|
|
|
|
Posix says you removing a non-empty directory "shall fail" in both:
https://pubs.opengroup.org/onlinepubs/9699919799.2018edition/functions/unlinkat.html
https://pubs.opengroup.org/onlinepubs/9699919799.2018edition/functions/rmdir.html
So toybox mv went ahead and unlinked the directory even if the contents
hadn't all been deleted because posix guarantees it to be harmless. But
cgroups (https://lwn.net/Articles/679786/) deletes the non-empty directory,
thus the Linux Test Project's cgroups_fj_function test6 was failing with
toybox mv because they depend on not triggering the posix violating behavior.
Work around it by having mv DIRTREE_SAVE failing nodes and then check for
a non-empty ->child in the COMEAGAIN as a signal not to unlink the dir.
While I'm there do some code cleanup, add a cp -i test...
|
