Skip to main content

Experimental extraction of Socorro signature generation

Project description

This is an experimental extraction of the Socorro signature generation code.

Code:

https://github.com/willkg/socorro-siggen

Documentation:

Check the README.rst file

Changelog:

Check the HISTORY.rst file

Issue tracker:

https://github.com/willkg/socorro-siggen/issues

License:

MPLv2

Status:

Alpha

Community Participation Guidelines:

Guidelines

Installing

socorro-siggen is available on PyPI. You can install it with:

$ pip install siggen

Basic use

Use it on the command line for signature generation debugging

You can generate a JSON output of the signature result from the command line:

$ signify <JSONFILE>
SIGNATURE HERE

Alternatively:

$ cat <JSONFILE> | signify
SIGNATURE HERE

You can pass in a --verbose flag and get verbose output about how the signature was generated.

You can generate a JSONFILE to pass into signify using crash ids from Crash Stats:

$ fetch-data CRASHID > crash.json

You can fetch crash data from Crash Stats, generate a signature, and get some output as to whether it matches the signature for that crash report in Crash Stats:

$ signature CRASHID

This helps when adjusting siglist and fixing signature generation problems.

Use it as a library

You can use socorro-siggen as a library:

from siggen.generator import SignatureGenerator

generator = SignatureGenerator()

crash_data = {
    ...
}

ret = generator.generate(crash_data)
print(ret['signature'])

Things to know

Things to know about siggen:

  1. Make sure to use the latest version of siggen and update frequently. We use semantic versioning, so the API will not change for MINOR and PATCH releases. Feel free to restrict on the MAJOR version.

  2. Signatures generated will change between siggen versions. The API may be stable, but bug fixes and changes to the siglist files will affect signature generation output. Hopefully for the better!

  3. If you have problems, please open up an issue. Please include the version of siggen.

    When using siggen, you can find the version like this:

    import siggen
    print(siggen.__version__)

Crash data schema

This is the schema for the crash data structure:

{
  crashing_thread: <int or null>,    // Optional, The index of the crashing thread in threads.
                                     // This defaults to None which indicates there was no
                                     // crashing thread identified in the crash report.

  threads: [                         // Optional, list of stack traces for c/c++/rust code.
    {
      frames: [                      // List of one or more frames.
        {
          function: <string>,        // Optional, The name of the function.
                                     // If this is ``None`` or not in the frame, then signature
                                     // generation will calculate something using other data in
                                     // the frame.

          module: <string>,          // Optional, name of the module
          file: <string>,            // Optional, name of the file
          line: <int>,               // Optional, line in the file
          module_offset: <string>,   // Optional, offset in hex in the module for this frame
          offset: <string>           // Optional, offset in hex for this frame

                                     // Signature parts are computed using frame data in this
                                     // order:

                                     // 1. if there's a function (and optionally line)--use
                                     //    that
                                     // 2. if there's a file and a line--use that
                                     // 3. if there's an offset and no module/module_offset--use
                                     //    that
                                     // 4. use module/module_offset
        }
        // ... additional frames
      ],

      thread_name: <string>,         // Optional, The name of the thread.
                                     // This isn't used, yet, but might be in the future for
                                     // debugging purposes.

      frame_count: <int>             // Optional, This is the total number of frames. This
                                     // isn't used.
    },
    // ... additional threads
  ],

  java_stack_trace: <string>,        // Optional, If the crash is a Java crash, then this will
                                     // be the Java traceback as a single string. Signature
                                     // generation will split this string into lines and then
                                     // extract frame information from it to generate the
                                     // signature.

                                     // FIXME(willkg): Write up better description of this.

  oom_allocation_size: <int>,        // Optional, The allocation size that triggered an
                                     // out-of-memory error. This will get added to the
                                     // signature if one of the indicator functions appears in
                                     // the stack of the crashing thread.

  abort_message: <string>,           // Optional, The abort message for the crash, if there is
                                     // one. This is added to the beginning of the signature.

  hang_type: <int>,                  // Optional.
                                     // 1 here indicates this is a chrome hang and we look at
                                     // thread 0 for generation.
                                     // -1 indicates another kind of hang.

  async_shutdown_timeout: <text>,    // Optional, This is a text field encoded in JSON with
                                     // "phase" and "conditions" keys.
                                     // FIXME(willkg): Document this structure better.

  jit_category: <string>,            // Optional, If there's a JIT classification in the
                                     // crash, then that will override the signature

  ipc_channel_error: <string>,       // Optional, If there is an IPC channel error, it
                                     // replaces the signature.

  ipc_message_name: <string>,        // Optional, This gets added to the signature if there
                                     // was an IPC message name in the crash.

  additional_minidumps: <string>,    // Optional, A crash report can contain multiple minidumps.
                                     // This is a comma-delimited list of minidumps other than
                                     // the main one that the crash had.

                                     // Example: "browser,flash1,flash2,content"

  mdsw_status_string: <string>,      // Optional, Socorro-generated
                                     // This is the minidump-stackwalk status string. This
                                     // gets generated when the Socorro processor runs the
                                     // minidump through minidump-stackwalk. If you're not
                                     // using minidump-stackwalk, you can ignore this.

  moz_crash_reason: <string>,        // Optional, This is the MOZ_CRASH_REASON value. This
                                     // doesn't affect anything unless the value is
                                     // "MOZ_RELEASE_ASSERT(parentBuildID == childBuildID)".

  os: <string>,                      // Optional, The name of the operating system. This
                                     // doesn't affect anything unless the name is "Windows
                                     // NT" in which case it will lowercase module names when
                                     // iterating through frames to build the signature.
}

Missing keys in the structure are treated as None, so you can pass in a minimal structure with just the parts you define.

Examples

Example almost minimal, somewhat nonsense crash_data.json:

{
    "os": "Linux",
    "crashing_thread": 0,
    "threads": [
        {
            "frames": [
                {
                    "frame": 0,
                    "function": "SomeFunc",
                    "line": 20,
                    "file": "somefile.cpp",
                    "module": "foo.so.5.15.0",
                    "module_offset": "0x37a92",
                    "offset": "0x7fc641052a92"
                },
                {
                    "frame": 1,
                    "function": "SomeOtherFunc",
                    "line": 444,
                    "file": "someotherfile.cpp",
                    "module": "bar.so",
                    "module_offset": "0x39a55",
                    "offset": "0x7fc641044a55"
                }
            ]
        }
    ]
}

That produces this output:

$ cat crash_data.json | signify
{
  "notes": [],
  "proto_signature": "SomeFunc | SomeOtherFunc",
  "signature": "SomeFunc"
}

Release process

  1. Create branch

  2. Update version and release date in siggen/__init__.py

  3. Update HISTORY.rst

  4. Push the branch, create a PR, review it, merge it

  5. Create a signed tag, push to github:

    git tag -s v0.1.0
    git push --tags REMOTE TAGNAME
  6. Build:

    python setup.py sdist bdist_wheel

    Make sure to use Python 3 with an updates requirements-dev.txt.

  7. Upload to PyPI:

    twine upload dist/*

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

siggen-1.0.2.tar.gz (53.5 kB view hashes)

Uploaded Source

Built Distribution

siggen-1.0.2-py3-none-any.whl (57.7 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page