commit mozjs24 for openSUSE:Factory

Hello community, here is the log from the commit of package mozjs24 for openSUSE:Factory checked in at 2016-07-01 09:51:39 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Comparing /work/SRC/openSUSE:Factory/mozjs24 (Old) and /work/SRC/openSUSE:Factory/.mozjs24.new (New) ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Package is "mozjs24" Changes: -------- --- /work/SRC/openSUSE:Factory/mozjs24/mozjs24.changes 2015-07-21 13:24:56.000000000 +0200 +++ /work/SRC/openSUSE:Factory/.mozjs24.new/mozjs24.changes 2016-07-01 09:51:41.000000000 +0200 @@ -1,0 +2,6 @@ +Mon Jun 13 17:13:12 UTC 2016 - agraf@suse.com + +- Fix 48 virtual address space systems like upstream (bsc#984126) + (mozjs-support-48bit-va.patch) + +------------------------------------------------------------------- New: ---- mozjs-support-48bit-va.patch ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ Other differences: ------------------ ++++++ mozjs24.spec ++++++ --- /var/tmp/diff_new_pack.UXJzkM/_old 2016-07-01 09:51:42.000000000 +0200 +++ /var/tmp/diff_new_pack.UXJzkM/_new 2016-07-01 09:51:42.000000000 +0200 @@ -1,7 +1,7 @@ # # spec file for package mozjs24 # -# Copyright (c) 2015 SUSE LINUX GmbH, Nuernberg, Germany. +# Copyright (c) 2016 SUSE LINUX GmbH, Nuernberg, Germany. # 2014 Wolfgang Rosenauer # # All modifications and additions to the file contributed by third parties @@ -28,6 +28,7 @@ Source1: baselibs.conf Patch1: mozjs-aarch64-support.patch Patch2: no_defined_array.patch +Patch3: mozjs-support-48bit-va.patch BuildRoot: %{_tmppath}/%{name}-%{version}-build BuildRequires: autoconf213 BuildRequires: gcc-c++ @@ -68,6 +69,7 @@ %setup -q -n mozjs-%{version} %patch1 -p1 %patch2 -p1 +%patch3 -p1 %build # no need to add build time to binaries ++++++ mozjs-support-48bit-va.patch ++++++ # HG changeset patch # User Zheng Xu # Date 1464657720 -7200 # Node ID dfaafbaaa2919a033c4c0abdd5830f4ea413bed6 # Parent 499f16ca85ec48d1896a1633730715f32bd62140 Bug 1143022 - Manually mmap on arm64 to ensure high 17 bits are clear. r=ehoogeveen There might be 48-bit VA on arm64 depending on kernel configuration. Manually mmap heap memory to align with the assumption made by JS engine. Index: mozjs-24.2.0/js/src/gc/Memory.cpp =================================================================== --- mozjs-24.2.0.orig/js/src/gc/Memory.cpp +++ mozjs-24.2.0/js/src/gc/Memory.cpp @@ -339,11 +339,44 @@ MapMemory(size_t length, int prot, int f return MAP_FAILED; } return region; +#elif defined(__aarch64__) + /* + * There might be similar virtual address issue on arm64 which depends on + * hardware and kernel configurations. But the work around is slightly + * different due to the different mmap behavior. + */ + const uintptr_t start = UINT64_C(0x0000070000000000); + const uintptr_t end = UINT64_C(0x0000800000000000); + const uintptr_t step = ChunkSize; + /* + * Optimization options if there are too many retries in practice: + * 1. Examine /proc/self/maps to find an available address. This file is + * not always available, however. In addition, even if we examine + * /proc/self/maps, we may still need to retry several times due to + * racing with other threads. + * 2. Use a global/static variable with lock to track the addresses we have + * allocated or tried. + */ + uintptr_t hint; + void* region = MAP_FAILED; + for (hint = start; region == MAP_FAILED && hint + length <= end; hint += step) { + region = mmap((void*)hint, length, prot, flags, fd, offset); + if (region != MAP_FAILED) { + if ((uintptr_t(region) + (length - 1)) & 0xffff800000000000) { + if (munmap(region, length)) { + MOZ_ASSERT(errno == ENOMEM); + } + region = MAP_FAILED; + } + } + } + return region == MAP_FAILED ? NULL : region; #else return mmap(NULL, length, prot, flags, fd, offset); #endif } + void * gc::MapAlignedPages(JSRuntime *rt, size_t size, size_t alignment) {