This commit fixes wchar.wcstold_hex_floats on arm64. On AArch64
(ARM64), the `long double` type has 128 bits and is more precise then
`double` type (64-bit). As a result, `1e100L` is slightly different
from `static_cast<long double>(1e100)`.
This commit fixes the regression by adding 'L' after the floating point
literals. This should work because casting from a higher precision
to lower precision won't lose any precisions.
Test: adb shell /data/nativetest64/bionic-unit-tests/bionic-unit-tests \
--gtest-filter=wchar.wcstold_hex_floats
Test: adb shell /data/nativetest/bionic-unit-tests/bionic-unit-tests
Test: adb shell /data/nativetest64/bionic-unit-tests/bionic-unit-tests
Change-Id: Ibd7b6a5d46c38338b4ca56838d9d272c710b32f6
...by inlining them.
Also fix a couple of harmless bugs in passing. I've added tests, but in
both cases I don't think it was actually possible to hit the bad behavior:
we'd hit another test and fail immediately after in an externally
indistinguishable way.
Bug: N/A
Test: readelf
Change-Id: I8466050b0bfe2b7b94c76b383cf10c1d9d28debd
The parsefloat routines -- which let us pass NaNs and infinities on to
strto(f|d|ld) -- come from NetBSD.
Also fix LP64's strtold to return a NaN, and fix all the architectures
to return quiet NaNs.
Also fix wcstof/wcstod/wcstold to use parsefloat so they support hex
floats.
Lots of new tests.
Bug: http://b/31101647
Change-Id: Id7d46ac2d8acb8770b5e8c445e87cfabfde6f111
Bionic never had this bug, but since the proposed fix is to remove the NDK's
broken code, we should add a regression test here.
Bug: https://code.google.com/p/android/issues/detail?id=80199
Change-Id: I4de21b5da9913cef990bc4d05a7e27562a71a02b
A mistake I made while cleaning this up the first time through.
mbstrtowcs(3) sets the src param to null if it finishes the string.
Change-Id: I6263646e25d9537043b7025fd1dd6ae195f365e2
The len parameter is a _maximum_ length. The previous code was treating
it as an exact length, causing the following typical call to fail:
mbsrtowcs(out, &in, sizeof(out), state); // sizeof(out) > strlen(in)
Change-Id: I48e474fd54ea5f122bc168a4d74bfe08704f28cc
The bug here turned out to be that we hadn't increased the constant
corresponding to the maximum number of bytes in a character to match
our new implementation, so any character requiring more than a byte
in UTF-8 would break our printf family.
Bug: 15439554
Change-Id: I693e5e6eb11c640b5886e848502908ec5fff53b1
Add optimized versions of bcopy and wmemmove for AArch64 based on the
memmove implementation
Change-Id: I82fbe8a7221ce224c567ffcfed7a94a53640fca8
Signed-off-by: Bernhard Rosenkraenzer <Bernhard.Rosenkranzer@linaro.org>
This reverts commit 8167dd7cb9.
For some reason I thought the bcopy change was bzero. The bcopy code doesn't pass our tests, so reverting until I can figure out what's wrong.
Change-Id: Id89fe959ea5105cd58dff6bba8d91a30cc4bcb07
Add optimized versions of bcopy and wmemmove for AArch64 based on the
memmove implementation
Change-Id: Ie43d0ff4f8ec4edba5b4fb5ccacd941f81ac6557
Signed-off-by: Bernhard Rosenkraenzer <Bernhard.Rosenkranzer@linaro.org>
I've also switched some tests to be positive rather than negative,
because !defined is slightly harder to reason about and there are
only two cases: bionic and glibc.
Change-Id: I8d3ac40420ca5aead3e88c69cf293f267273c8ef
when compile the cts package with aarch64 gcc4.9, will get following error:
bionic/tests/wchar_test.cpp:253:3: required from here
external/gtest/include/gtest/gtest.h:1448:16:
error: comparison between signed and unsigned integer expressions [-Werror=sign-compare]
this change fix it by using static_cast<wchar_t> as suggested by Calin Juravle
Change-Id: I7fb9506e7b84b8a12b9d003458d4f0e78554c3cd
Signed-off-by: Yongqin Liu <yongqin.liu@linaro.org>
Although glibc gets by with an 8-byte mbstate_t, OpenBSD uses 12 bytes (of
the 128 bytes it reserves!).
We can actually implement UTF-8 encoding/decoding with a 0-byte mbstate_t
which means we can make things work on LP32 too, as long as we accept the
limitation that the caller needs to present us with a complete sequence
before we'll process it.
Our behavior is fine when going from characters to bytes; we just
update the source wchar_t** to say how far through the input we got.
I'll come back and use the 4 bytes we do have to cope with byte sequences
split across multiple input buffers. The fact that we don't support
UTF-8 sequences longer than 4 bytes plus the fact that the first byte of
a UTF-8 sequence encodes the length means we shouldn't need the other
fields OpenBSD used (at the cost of some recomputation in cases where a
sequence is split across buffers).
This patch also makes the minimal changes necessary to setlocale(3) to
make us behave like glibc when an app requests UTF-8. (The difference
being that our "C" locale is the same as our "C.UTF-8" locale.)
Change-Id: Ied327a8c4643744b3611bf6bb005a9b389ba4c2f
This replaces a partial set of non-functional functions with a complete
set of functions, all of which actually work.
This requires us to implement mbsnrtowcs and wcsnrtombs which completes
the set of what we need for libc++.
The mbsnrtowcs is basically a copy & paste of wcsnrtombs, but I'm going
to go straight to looking at using the OpenBSD UTF-8 implementation rather
than keep polishing our home-grown turd.
(This patch also opportunistically switches us over to upstream btowc,
mbrlen, and wctob, since they're all trivially expressed in terms of
other functions.)
Change-Id: I0f81443840de0f1aa73b96f0b51988976793a323
This is an implementation in the style of the rest: char == byte.
We might want to come back and implement UTF-8, but this is enough for ltrace.
Bug: 13747066
Change-Id: Ib2b63609c9014fdef9a8491e067467c4fc5ae3cc
