)]}'
{
  "log": [
    {
      "commit": "d7c52625f2da8322b4f0adb4c67674f0d66e619a",
      "tree": "a885e0530cb1a26a5a1fd0dee7477ef02dcb4833",
      "parents": [
        "b1b7eda11e4f66b608ae66945ddf87cbe31ecb47"
      ],
      "author": {
        "name": "Elliott Hughes",
        "email": "enh@google.com",
        "time": "Tue Jun 20 14:26:56 2017 -0700"
      },
      "committer": {
        "name": "Elliott Hughes",
        "email": "enh@google.com",
        "time": "Tue Jun 20 14:58:16 2017 -0700"
      },
      "message": "Shave another uninteresting stack frame off aborts.\n\nWith this, stack frame 0 is the abort, not tgkill.\n\narm:\n\n     #00 pc 0001a41c  /system/lib/libc.so (abort+63)\n\narm64:\n\n     #00 pc 000000000001d75c  /system/lib64/libc.so (abort+120)\n\nAlso \"include what you use\" for \u003csys/syscall.h\u003e.\n\nBug: N/A\nTest: ran `crasher abort` and `crasher64 abort`\nChange-Id: I6517ac67b39b4133e890d52efc115071c812958b\n"
    },
    {
      "commit": "bf2af69fb2ff1365f4075d0207d6a895bc2727ed",
      "tree": "f4ff284b54aef4d14c18ecb3d355b81bb5878c6a",
      "parents": [
        "05fd20c797c0aab700183131023130e45a748a47"
      ],
      "author": {
        "name": "Josh Gao",
        "email": "jmgao@google.com",
        "time": "Fri May 05 13:27:15 2017 -0700"
      },
      "committer": {
        "name": "Josh Gao",
        "email": "jmgao@google.com",
        "time": "Fri May 05 14:25:24 2017 -0700"
      },
      "message": "Make raise/abort work with stale cached pid/tid values.\n\nSwitch raise to using tgkill with direct syscalls of getpid/gettid,\nand switch abort to use raise(SIGABRT).\n\nBug: http://b/37769298\nTest: debuggerd_test\nChange-Id: If6f9d17fd8ae6177e742dc9f2f44bd78539431ba\n"
    },
    {
      "commit": "d3cfd2687249d2340b738ddeb8ea0660860e9e3f",
      "tree": "02bc3a8cac4de0b2a7059636cb92cf0d7d953ac7",
      "parents": [
        "fb07c36bc061db4ca5d8348ff6bc1e60b6c53191"
      ],
      "author": {
        "name": "Josh Gao",
        "email": "jmgao@google.com",
        "time": "Thu Jan 05 13:22:26 2017 -0800"
      },
      "committer": {
        "name": "Josh Gao",
        "email": "jmgao@google.com",
        "time": "Thu Jan 05 15:15:12 2017 -0800"
      },
      "message": "Add declaration of tgkill to signal.h.\n\nExpose a useful function that we\u0027ve had since Jelly Bean.\n\nBug: http://b/34111810\nTest: TreeHugger\nChange-Id: Iaf3097f224c09b533f36050cf21394ba148007ad\n"
    },
    {
      "commit": "4bf55226623801fab4751bc0736a07e044d00b49",
      "tree": "38365d7202df1459e15a0eae01a1e1144f0177ea",
      "parents": [
        "bd734f07167b9e24cd292e669717795f83d10157"
      ],
      "author": {
        "name": "Elliott Hughes",
        "email": "enh@google.com",
        "time": "Tue Nov 29 16:59:26 2016 -0800"
      },
      "committer": {
        "name": "Elliott Hughes",
        "email": "enh@google.com",
        "time": "Tue Nov 29 22:02:25 2016 -0800"
      },
      "message": "Abort harder.\n\nSome PoS internal system can\u0027t cope with more than 4 stack frames,\nso the fact that our abort(3) implementation takes 4 frames by itself\nmakes it useless.\n\nRe-reading POSIX, it only says \"behaves as if\", so the previous\nimplementation chain wasn\u0027t mandatory and we can just go straight to\ncalling tgkill...\n\nBefore:\n\n     #00 pc 0000000000069be4  /system/lib64/libc.so (tgkill+8)\n     #01 pc 0000000000066d50  /system/lib64/libc.so (pthread_kill+64)\n     #02 pc 0000000000028110  /system/lib64/libc.so (raise+24)\n     #03 pc 000000000001d4ec  /system/lib64/libc.so (abort+52)\n\nAfter:\n\n     #00 pc 0000000000069bc8  /system/lib64/libc.so (tgkill+8)\n     #01 pc 000000000001d4c8  /system/lib64/libc.so (abort+80)\n     #02 pc 0000000000001494  /system/xbin/crasher64 (_ZL9do_actionPKc+872)\n     #03 pc 00000000000010e0  /system/xbin/crasher64 (main+88)\n\nThis is less useful on 32-bit ARM because there there\u0027s an extra trampoline\nfrom an assembler abort(3) implementation, so you\u0027ll still only get one\nmeaningful stack frame. But every other architecture will now get two!\n\nBut wait!\n\nIt turns out that the assembler hack isn\u0027t needed any more. Here we are\nunwinding just fine all the way through the 32-bit ARM crasher:\n\nBefore (with direct call to tgkill but still using the assembler):\n\n     #00 pc 00049e7c  /system/lib/libc.so (tgkill+12)\n     #01 pc 00019c6f  /system/lib/libc.so (__libc_android_abort+50)\n     #02 pc 000181f8  /system/lib/libc.so (abort+4)\n     #03 pc 00001025  /system/xbin/crasher (_ZL9do_actionPKc+656)\n     #04 pc 00017721  /system/lib/libc.so (__libc_init+48)\n     #05 pc 00000b38  /system/xbin/crasher (_start+96)\n\nAfter:\n\n     #00 pc 00049e6c  /system/lib/libc.so (tgkill+12)\n     #01 pc 00019c5f  /system/lib/libc.so (abort+50)\n     #02 pc 00001025  /system/xbin/crasher (_ZL9do_actionPKc+656)\n     #03 pc 00017721  /system/lib/libc.so (__libc_init+48)\n     #04 pc 00000b38  /system/xbin/crasher (_start+96)\n\n(As you can see, the fact that we see __libc_init rather than main was true\nwith the assembler stub too, so that\u0027s not a regression even if it does seem\nodd...)\n\nBug: N/A\nTest: ran crasher64\nChange-Id: I9dd5b214c495604c8b502c7ec0de3631080d8c29\n"
    },
    {
      "commit": "53c3c271dc9927dd280981fc23409af60f460007",
      "tree": "4a04122167583b9994204372bbc549a8174b9611",
      "parents": [
        "83ce99d8b7e2b15b2169e6e6e18e871ad35abb6c"
      ],
      "author": {
        "name": "Dmitriy Ivanov",
        "email": "dimitry@google.com",
        "time": "Fri Jul 11 12:59:16 2014 -0700"
      },
      "committer": {
        "name": "Dmitriy Ivanov",
        "email": "dimitry@google.com",
        "time": "Mon Jul 14 12:05:16 2014 -0700"
      },
      "message": "Upstream atexit\n\nChange-Id: Ia454a2181b5058ed9783dc02b6b1805d0e4d2715\n"
    },
    {
      "commit": "623b0d05bd924e7fa3c155492695123c87c3a2b1",
      "tree": "8fa47b2bf2fb7abe4b168fb1009108cdd762f6e3",
      "parents": [
        "089c7c921786ff6255526feec96552fc16ada7f4"
      ],
      "author": {
        "name": "Dmitriy Ivanov",
        "email": "dimitry@google.com",
        "time": "Wed May 14 23:11:05 2014 -0700"
      },
      "committer": {
        "name": "Dmitriy Ivanov",
        "email": "dimitry@google.com",
        "time": "Thu May 15 13:05:21 2014 -0700"
      },
      "message": "Register _cleanup function with atexit\n\n * Register cleanup function with atexit\n   instead of calling it explicitly on\n   exit()\n * abort() no longer calls _cleanup:\n   Flushing stdio buffers on abort is no\n   longer required by POSIX.\n * dlmalloc no longer need to reset cleanup\n   (see above)\n * Upstream findfp.c makebuf.c setvbuf.cexit.c\n   to openbsd versions.\n\nBug: 14415367\nChange-Id: I277058852485a9d3dbb13e5c232db5f9948d78ac\n"
    },
    {
      "commit": "61e699a133a4807fe878a6cb0d7190d7c96e21f8",
      "tree": "12dd46e7f8e5f7172a3fe0fd9075fa929cbb1001",
      "parents": [
        "5cde15eb17f4ddc0949c0cbd38a46e6c0bad7019"
      ],
      "author": {
        "name": "Elliott Hughes",
        "email": "enh@google.com",
        "time": "Wed Jun 12 14:05:46 2013 -0700"
      },
      "committer": {
        "name": "Elliott Hughes",
        "email": "enh@google.com",
        "time": "Wed Jun 12 14:14:53 2013 -0700"
      },
      "message": "Clean up abort.\n\n* A dlmalloc usage error shouldn\u0027t call abort(3) because we want to\n  cause a SIGSEGV by writing the address dlmalloc didn\u0027t like to an\n  address the kernel won\u0027t like, so that debuggerd will dump the\n  memory around the address that upset dlmalloc.\n\n* Switch to the simpler FreeBSD/NetBSD style of registering stdio\n  cleanup. Hopefully this will let us simplify more of the stdio\n  implementation.\n\n* Clear the stdio cleanup handler before we abort because of a dlmalloc\n  corruption error. This fixes the reported bug, where we\u0027d hang inside\n  dlmalloc because the stdio cleanup reentered dlmalloc.\n\nBug: 9301265\nChange-Id: Ief31b389455d6876e5a68f0f5429567d37277dbc\n"
    }
  ]
}