https://sources.debian.org/data/main/a/abseil/0~20200923.3-2/debian/patches/cpu-frequency.diff This patch is taken from Debian instead of the upstream URL because the upstream URL contains far more changes than occur in this patch. It was then modified to also work for armhf. From: Benjamin Barenblat Subject: Ignore missing CPU frequency on more architectures Forwarded: yes Applied-Upstream: https://github.com/abseil/abseil-cpp/commit/1918ad2ae38aa32c74b558b322479a8efdd76363 Linux on MIPS, PA-RISC, RISC-V, and SystemZ doesn’t expose the nominal CPU frequency via /sys, so don’t worry if `NominalCPUFrequency` returns 1.0 on those platforms. Some POWER machines expose the CPU frequency; others do not. Since we can’t predict which type of machine the tests will run on, simply disable testing for `NominalCPUFrequency` on POWER. The author works at Google. Upstream applied this patch as Piper revision 347079873 and exported it to GitHub; the Applied-Upstream URL above points to the exported commit. --- a/absl/base/internal/sysinfo_test.cc +++ b/absl/base/internal/sysinfo_test.cc @@ -37,17 +37,28 @@ TEST(SysinfoTest, NumCPUs) { << "NumCPUs() should not have the default value of 0"; } +// Ensure that NominalCPUFrequency returns a reasonable value, or 1.00 on +// platforms where the CPU frequency is not available through sysfs. +// +// POWER is particularly problematic here; some Linux kernels expose the CPU +// frequency, while others do not. Since we can't predict a priori what a given +// machine is going to do, just disable this test on POWER on Linux. +#if !(defined(__linux) && (defined(__ppc64__) || defined(__PPC64__))) TEST(SysinfoTest, NominalCPUFrequency) { -#if !(defined(__aarch64__) && defined(__linux__)) && !defined(__EMSCRIPTEN__) - EXPECT_GE(NominalCPUFrequency(), 1000.0) - << "NominalCPUFrequency() did not return a reasonable value"; -#else - // Aarch64 cannot read the CPU frequency from sysfs, so we get back 1.0. - // Emscripten does not have a sysfs to read from at all. + // Linux only exposes the CPU frequency on certain architectures, and + // Emscripten doesn't expose it at all. +#if defined(__linux__) && \ + (defined(__aarch64__) || defined(__hppa__) || defined(__mips__) || \ + defined(__arm__) || defined(__riscv) || defined(__s390x__)) || \ + defined(__EMSCRIPTEN__) EXPECT_EQ(NominalCPUFrequency(), 1.0) << "CPU frequency detection was fixed! Please update unittest."; +#else + EXPECT_GE(NominalCPUFrequency(), 1000.0) + << "NominalCPUFrequency() did not return a reasonable value"; #endif } +#endif TEST(SysinfoTest, GetTID) { EXPECT_EQ(GetTID(), GetTID()); // Basic compile and equality test. etc/committer.scm.in (change-commit-message): Use object->string to deal with more complex dependency specifications. Change-Id: Ieee39709fd3eb26368df400a9228d77aa97a9d9d Signed-off-by: Ludovic Courtès <ludo@gnu.org> Attila Lendvai 2024-08-26committer: Avoid // in xpath....When using // we assume that the package definition's S-expression is a nesting of proper lists. This is not guaranteed to be true, because a package definition may contain alists. By using an explicit path to the field of interest we avoid recursion through all child nodes, thus avoiding child nodes that are more likely to contain improper lists. * etc/committer.scm.in (change-commit-message): Replace // with a path for the PACKAGE symbol. Change-Id: Idefaec9f5c3538484432e15a5865a68fc16e7369 Ricardo Wurmus