Export of internal Abseil changes.

--
898e99cae89ca4cc27f86f719148f020d521dd58 by Abseil Team <absl-team@google.com>:

Update comment.

PiperOrigin-RevId: 204323401

--
b9d14db8b8a9dfb0e1cfb5967aaa0de1c4e94c42 by Abseil Team <absl-team@google.com>:

Internal change

PiperOrigin-RevId: 204178059

--
f3b5a667611a588aa06fea9168e997ef5cffa7ac by Abseil Team <absl-team@google.com>:

Fix a potential reinterpret_cast compile error in absl::InlinedVector

The current code will trigger a reinterpret_cast error enhanced by llvm r336738.

PiperOrigin-RevId: 204131536

--
cc87d7a8302ad4471c1a25781d6ec19c2ce1524e by Abseil Team <absl-team@google.com>:

Internal change.

PiperOrigin-RevId: 203979040

--
bc5cae3cfc72af1d3e786d5a3b59a47e205afec9 by Gennadiy Rozental <rogeeff@google.com>:

Internal: add internal logging hooks

PiperOrigin-RevId: 203850605

--
503655c248f557f677c920078613522b010e73c8 by Derek Mauro <dmauro@google.com>:

Cleanup stacktrace_config.h

Instead of listing the platforms that aren't supported, list the ones
we do support, and fallback to stacktrace_unimplemented-inl.inc at the
end. Previously any platform that wasn't listed gets "#error Not supported yet".

GitHub issue #135

PiperOrigin-RevId: 203823079

--
cb69925c4cacc14558cf103a09f218c37f466a16 by Abseil Team <absl-team@google.com>:

Fix a minor typo in absl::variant documentation.

PiperOrigin-RevId: 203679877

--
23a0e4db10039011fa5fd879fb73d2f2bbd17301 by Abseil Team <absl-team@google.com>:

Format .bzl files with buildifier

PiperOrigin-RevId: 203461813

--
1ad02616bdb715dfdc7f1a73313e383f4ededa03 by Abseil Team <absl-team@google.com>:

Make the absl::SleepFor() tests treat any successful retry within
a 48x deadline as a total success, thereby reducing flakiness.

PiperOrigin-RevId: 203401603

--
b3dccbbc6563ea0173527076a3fff21433d48f47 by Abseil Team <absl-team@google.com>:

Replace config_setting.values{"compiler"} with config_setting.flag_values{"@bazel_tools//tools/cpp:compiler"}

Due to changes in Bazel we need to change the way "compiler" is specified in config_setting. This will not change the behavior of the config_setting itself.

PiperOrigin-RevId: 203345693

--
170f1692537460a4ba1756d34852275899c2339b by Matt Armstrong <marmstrong@google.com>:

Address test flakiness in the TimeoutTest.

The basic idea is to loop forever until the expected
scheduling delays are observed (within reasonable bounds),
and only then assert the other invariants.

PiperOrigin-RevId: 203188162
GitOrigin-RevId: 898e99cae89ca4cc27f86f719148f020d521dd58
Change-Id: Ie853ec050afa3a04c519393afe666bc03e11dc87

1 files changed, 2 insertions, 2 deletions

diff --git a/absl/strings/str_format_test.cc b/absl/strings/str_format_test.cc
index fe742bf9..fed75faf 100644
--- a/absl/strings/str_format_test.cc
+++ b/absl/strings/str_format_test.cc
@@ -384,8 +384,8 @@ TEST(StrFormat, BehavesAsDocumented) {
   EXPECT_EQ(StrFormat("%G", 1e10), "1E+10");
   //     a/A - lower,upper case hex    Eg: -3.0 -> "-0x1.8p+1"/"-0X1.8P+1"
 
-// On NDK r16, there is a regression in hexfloat formatting.
-#if !defined(__NDK_MAJOR__) || __NDK_MAJOR__ != 16
+// On Android platform <=21, there is a regression in hexfloat formatting.
+#if !defined(__ANDROID_API__) || __ANDROID_API__ > 21
   EXPECT_EQ(StrFormat("%.1a", -3.0), "-0x1.8p+1");  // .1 to fix MSVC output
   EXPECT_EQ(StrFormat("%.1A", -3.0), "-0X1.8P+1");  // .1 to fix MSVC output
 #endif