{"full_name":"gperftools/gperftools","owner":"gperftools","name":"gperftools","shard":"1","url":"https://github.com/gperftools/gperftools","description":"Main gperftools repository","homepage":"","topics":[],"tag_pages":[],"language":"C++","license":"BSD-3-Clause","avatar":"https://avatars.githubusercontent.com/u/6523841?v=4","created_at":"2014-01-28T11:15:07Z","pushed_at":"2026-11-14T02:00:15Z","metrics":{"stars":8989,"forks":1537,"watchers":353,"issues":1383,"prs":260,"releases":48,"issues_open":103,"issues_closed":1280,"prs_open":4,"prs_closed":200,"prs_merged":56,"discussions":10},"growth":{"day":{"stars":1,"forks":0,"base_day":"2026-11-15"}},"history":{"dates":["2026-11-10","2026-11-11","2026-11-12","2026-11-13","2026-11-14","2026-11-15","2026-11-16"],"stars":[8987,8988,8987,8987,8987,8988,8989],"forks":[1536,1536,1536,1537,1537,1537,1537],"watchers":[null,353,353,353,353,353,353],"issues_open":[null,105,105,105,104,103,103],"prs_open":[null,4,4,5,4,4,4]},"boards":{},"same_owner":[],"issues":[{"number":1066,"title":"dynamically loading libtcmalloc is unsafe (was: Dynamic linking with tcmalloc is prone to crash)","comments":21,"created_at":"2018-11-05T04:12:42Z"},{"number":1538,"title":"memfs sysallocator can't share fd across forking (was: SIGBUS error in child after parent performs allocation on Hugepages)","comments":18,"created_at":"2024-07-24T08:54:34Z"},{"number":270,"title":"Some issues of HeapProfile in windows","comments":16,"created_at":"2015-08-22T22:56:26Z"},{"number":1318,"title":"build for Android/x86_64 failed","comments":15,"created_at":"2021-12-24T10:04:15Z"},{"number":724,"title":"8 bytes size class in tcmalloc and jemalloc breaks ABI on x86_64","comments":13,"created_at":"2015-09-30T15:54:50Z"},{"number":1167,"title":"Insufficient control over RUNNING_IN_VALGRIND, should only be active with --tool=memcheck","comments":12,"created_at":"2020-01-23T21:33:08Z"},{"number":758,"title":"valgrind reports that tcmalloc from gperftools-2.4 leaks some memory","comments":12,"created_at":"2016-01-14T09:32:20Z"},{"number":1576,"title":"CMake scripts has no install() directives","comments":9,"created_at":"2024-12-13T11:23:08Z"}],"discussions":[{"number":1654,"title":"gperftools-2.19.3","comments":0,"category":"Announcements","created_at":"2026-11-10T18:19:58Z"},{"number":1651,"title":"gperftools-2.19.2","comments":0,"category":"Announcements","created_at":"2026-10-29T02:18:19Z"},{"number":1648,"title":"gperftools-2.19","comments":0,"category":"Announcements","created_at":"2026-10-20T17:04:30Z"},{"number":1646,"title":"gperftools-2.18.91","comments":0,"category":"Announcements","created_at":"2026-10-01T19:31:05Z"},{"number":1643,"title":"gperftools-2.18.90","comments":0,"category":"Announcements","created_at":"2026-09-05T22:46:34Z"},{"number":1632,"title":"gperftools-2.18.1","comments":0,"category":"Announcements","created_at":"2026-03-06T20:34:34Z"},{"number":1627,"title":"gperftools-2.18","comments":0,"category":"Announcements","created_at":"2026-01-25T18:29:19Z"},{"number":1605,"title":"gperftools-2.17","comments":0,"category":"Announcements","created_at":"2025-08-04T21:31:19Z"}],"latest_release":{"name":"gperftools-2.19.3","tag":"gperftools-2.19.3","published_at":"2026-11-10T18:19:58Z","body":"This is another bug-fix release on top of 2.19.2. It updates the vendored aw-backtrace.\r\n\r\nWith the previous version of aw-backtrace, we got broken backtraces on aarch64 when gperftools was built with -fstack-protector-strong, which is the default compiler setting on many distros, e.g. Ubuntu (see https://github.com/gperftools/gperftools/issues/1653). This release fixes that and related issues. aw-backtrace's own CI now also tests builds with stack protector, hardening (-mbranch-protection or -fcf-protection), without frame pointers, and, on x86-64, x86-64-v3 builds (stack realignment).\r\n\r\nNotable changes:\r\n\r\n* aw-backtrace now has a small piece of assembly (aw-asm-x86_64.S and aw-asm-aarch64.S) that captures the exact pc, sp and frame pointer of aw_backtrace_full's own frame, so the backtrace starts from known values. It is used on both architectures. Previously, on aarch64, we guessed the stack pointer of the frame we started from, and that guess was the bug described above.\r\n* aarch64: when a frame has no usable unwind info, aw-backtrace steps to the caller through the frame record (saved x29/x30). A frame record gives the caller's pc and frame pointer, but not its stack pointer. We used to guess the stack pointer; now it is treated as unknown until the caller's own unwind info makes it known again.\r\n* aw-backtrace now steps through frames whose unwind info uses an unsupported DWARF expression via the fallback (frame record based on aarch64), instead of stopping the backtrace there. This fixes backtraces on aarch64 through functions with SVE stack slots.\r\n* aarch64: aw-backtrace now understands the DW_CFA_AARCH64_negate_ra_state_with_pc unwind instruction, which clang emits with -mbranch-protection=pac-ret+pc. Previously, frames using it were handled by the fallback instead of their unwind info.\r\n* the Bazel build now pulls aw-backtrace via git_override, since the needed version isn't in the BCR yet.\r\n","is_prerelease":false},"readme":false,"readme_zh":false,"news":[],"stale":false,"snapshot_at":"2026-11-16T09:23:14Z","history_start":"2026-09-06"}