Skip to main content

A parser for GDB Machine Interface (MI) events.

Project description

Build Status

Python GDB MI Parser

MI or Machine Interface is the new interface to interact with GDB, the GNU Debugger, from another program.

The output of the GDB Machine Interface is line oriented, text based. It is compound of small elements that range from strings to dictionaries

python-gdb-mi is simple and quite robust parser for Python 2.x/3.x that can take those lines and transform them into python objects ready to be serialized if need to JSON.

Overview

A GDB MI text can be like this:

>>> text = '^done,bkpt={number="1",type="breakpoint",disp="keep",enabled="y",addr="0x08048564",func="main",file="myprog.c",fullname="/home/nickrob/myprog.c",line="68",thread-groups=["i1"],times="0"}\n'

This is the kind of message that GDB will print when a breakpoint is set.

To parse it, we need to send this line to our Output parser using the parse_line method:

>>> from gdb_mi import Output

>>> out = Output()

>>> record = out.parse_line(text)
>>> record                                #doctest: +NORMALIZE_WHITESPACE
{'klass': 'done',
 'results': {'bkpts': [{'addr': '0x08048564',
                        'disp': 'keep',
                        'enabled': 'y',
                        'file': 'myprog.c',
                        'fullname': '/home/nickrob/myprog.c',
                        'func': 'main',
                        'line': '68',
                        'number': '1',
                        'thread-groups': ['i1'],
                        'times': '0',
                        'type': 'breakpoint'}]},
 'token': None,
 'type': 'Sync'}

If the output from GDB is not a complete line, Output can handle it anyways doing some buffering. Use parse instead of parse_line to feed Output:

>>> out.parse(text[:10])     # incomplete line, None returned

>>> out.parse(text[10:])     # enough data, parse it! doctest: +NORMALIZE_WHITESPACE
{'klass': 'done',
 'results': {'bkpts': [{'addr': '0x08048564',
                        'disp': 'keep',
                        'enabled': 'y',
                        'file': 'myprog.c',
                        'fullname': '/home/nickrob/myprog.c',
                        'func': 'main',
                        'line': '68',
                        'number': '1',
                        'thread-groups': ['i1'],
                        'times': '0',
                        'type': 'breakpoint'}]},
 'token': None,
 'type': 'Sync'}

Parsing Results

Three types of objects can be returned by parse_line and parse:

  • Stream that represents an output record from: the console, the target and the log.
  • Record that represents or a synchronous result record and or an out of band asynchronous record, used to notify of changes that have happen.
  • (gdb) a literal string that represents an empty prompt line.

Both, Stream and Record have a as_native method to transform them into a composition of Python’s dicts and lists.

Streams

>>> from gdb_mi import Stream

>>> text = '~"GDB rocks!"\n'
>>> stream = out.parse_line(text)
>>> stream      # same as pprint.pprint(stream.as_native())
{'stream': 'GDB rocks!', 'type': 'Console'}

>>> isinstance(stream, Stream)
True

The type attribute is one of the following, from the GDB MI’s documentation: - Console: output that should be displayed as is in the console. It is the textual response to a CLI command. - Target: output produced by the target program. - Log: output text coming from GDB’s internals, for instance messages that should be displayed as part of an error log.

Records

We have already seen an example of a Record, in that case it was a synchronous result record:

>>> from gdb_mi import Record

>>> isinstance(record, Record)
True

>>> record.klass, record.type
('done', 'Sync')

The klass attribute is one of the following: done, running, connected, error or exit.

The type attribute is Sync for a synchronous result record.

Here are an example of an asynchronous record:

>>> text = '42*stopped,reason="breakpoint-hit",disp="keep",bkptno="1",thread-id="0",frame={addr="0x08048564",func="main",args=[{name="argc",value="1"},{name="argv",value="0xbfc4d4d4"}],file="myprog.c",fullname="/home/nickrob/myprog.c",line="68"}\n'
>>> record = out.parse_line(text)
>>> record                                #doctest: +NORMALIZE_WHITESPACE
{'klass': 'stopped',
  'results': {'bkptno': '1',
              'disp': 'keep',
              'frame': {'addr': '0x08048564',
                        'args': [{'name': 'argc', 'value': '1'},
                                 {'name': 'argv', 'value': '0xbfc4d4d4'}],
                        'file': 'myprog.c',
                        'fullname': '/home/nickrob/myprog.c',
                        'func': 'main',
                        'line': '68'},
              'reason': 'breakpoint-hit',
              'thread-id': '0'},
  'token': 42,
  'type': 'Exec'}

>>> isinstance(record, Record)
True

>>> record.klass, record.type
('stopped', 'Exec')

For an asynchronous record, the attribute type is one of the following for AsyncRecords: Exec, Status or Notify.

>From the GDB MI’s documentation: - Exec: asynchronous state change on the target (stopped, started, disappeared). - Status: on-going status information about the progress of a slow operation. It can be discarded. - Notify: supplementary information that the client should handle (e.g., a new breakpoint information).

Both kind of records, synchronous and asynchronous, have two additional attributes: - token: used by GDB to match the request and the response. - results: the data contained in the message, it will depend of the GDB message.

Interference from Target

If you do not redirect the target’s output nor send it to a new console running the GDB set new-console on command, the output of the target will interfere an confuse the parser.

Unfortunately there is nothing that we can do. Even if we ignore the message we cannot be sure when a message is safe to be discarded.

For example, the following C code generates an ambiguous output:

printf("~looks like a GDB stream but it isn't\n");

Even if you think that it is improbable, here is a quite common problem:

printf("normal output 42"); /* no newline at the end */
fflush(stdout); /* but we flush to the console anyway */

Now imagine that GDB hits a breakpoint after the fflush instruction, what we will see is:

>>> text = 'normal output 4242*stopped,reason="breakpoint-hit",<and so on...>\n'

The problem is that all those strings are glued together which can lead to nasty bugs. We could try to use some regexps but it would be too fragile (is the token 42 or 4242?).

Instead we try to warn you if you try to parse something like that:

>>> out.parse_line(text)                     #doctest: +IGNORE_EXCEPTION_DETAIL
Traceback (most recent call last):
ParsingError: Invalid input. Maybe the target's output is interfering with the GDB MI's messages. Try to redirect the target's output to elsewhere or run GDB's 'set new-console on' command. Found at 0 position.
Original message:
  normal output 4242*stopped,reason="breakpoint-hit",<and so on...>

Install

Just run:

$ pip install python-gdb-mi

You will find the python-gdb-mi package at PyPI

Workarounds for GDB MI’s issues

There are some issues in the output of GDB. python-gdb-mi tries to fix them implementing some minor changes in the GDB’s output as workarounds.

See the issues and the implemented fixes in the workarounds doctest

Hacking/Contributing

Go ahead! Clone the repository, do a small fix/enhancement, run make test to ensure that everything is working as expected and then propose your Pull Request!

Project details


Download files

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

Filename, size & hash SHA256 hash help File type Python version Upload date
python_gdb_mi-1.0.1-py2.py3-none-any.whl (11.7 kB) Copy SHA256 hash SHA256 Wheel py2.py3
python-gdb-mi-1.0.1.tar.gz (8.0 kB) Copy SHA256 hash SHA256 Source None

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN SignalFx SignalFx Supporter DigiCert DigiCert EV certificate StatusPage StatusPage Status page