cborutil: implement sans I/O decoder
The vendored CBOR package decodes by calling read(n) on an object.
There are a number of disadvantages to this:
* Uses blocking I/O. If sufficient data is not available, the decoder
will hang until it is.
* No support for partial reads. If the read(n) returns less data than
requested, the decoder raises an error.
* Requires the use of a file like object. If the original data is in
say a buffer, we need to "cast" it to e.g. a BytesIO to appease the
decoder.
In addition, the vendored CBOR decoder doesn't provide flexibility
that we desire. Specifically:
* It buffers indefinite length bytestrings instead of streaming them.
* It doesn't allow limiting the set of types that can be decoded. This
property is useful when implementing a "hardened" decoder that is
less susceptible to abusive input.
* It doesn't provide sufficient "hook points" and introspection to
institute checks around behavior. These are useful for implementing
a "hardened" decoder.
This all adds up to a reasonable set of justifications for writing our
own decoder.
So, this commit implements our own CBOR decoder.
At the heart of the decoder is a function that decodes a single "item"
from a buffer. This item can be a complete simple value or a special
value, such as "start of array." Using this function, we can build a
decoder that effectively iterates over the stream of decoded items and
builds up higher-level values, such as arrays, maps, sets, and indefinite
length bytestrings. And we can do this without performing I/O in the
decoder itself.
The core of the sans I/O decoder will probably not be used directly.
Instead, it is expected that we'll build utility functions for invoking
the decoder given specific input types. This will allow extreme
flexibility in how data is delivered to the decoder.
I'm pretty happy with the state of the decoder modulo the TODO items
to track wanted features to help with a "hardened" decoder. The one
thing I could be convinced to change is the handling of semantic tags.
Since we only support a single semantic tag (sets), I thought it would
be easier to handle them inline in decodeitem(). This is simpler now.
But if we add support for other semantic tags, it will likely be easier
to move semantic tag handling outside of decodeitem(). But, properly
supporting semantic tags opens up a whole can of worms, as many
semantic tags imply new types. I'm optimistic we won't need these in
Mercurial. But who knows.
I'm also pretty happy with the test coverage. Writing comprehensive
tests for partial decoding did flush out a handful of bugs. One
general improvement to testing would be fuzz testing for partial
decoding. I may implement that later. I also anticipate switching the
wire protocol code to this new decoder will flush out any lingering
bugs.
Differential Revision: https://phab.mercurial-scm.org/D4414
Test UI worker interaction
$ cat > t.py <<EOF
> from __future__ import absolute_import, print_function
> import time
> from mercurial import (
> error,
> registrar,
> ui as uimod,
> worker,
> )
> def abort(ui, args):
> if args[0] == 0:
> # by first worker for test stability
> raise error.Abort(b'known exception')
> return runme(ui, [])
> def exc(ui, args):
> if args[0] == 0:
> # by first worker for test stability
> raise Exception('unknown exception')
> return runme(ui, [])
> def runme(ui, args):
> for arg in args:
> ui.status(b'run\n')
> yield 1, arg
> time.sleep(0.1) # easier to trigger killworkers code path
> functable = {
> b'abort': abort,
> b'exc': exc,
> b'runme': runme,
> }
> cmdtable = {}
> command = registrar.command(cmdtable)
> @command(b'test', [], b'hg test [COST] [FUNC]')
> def t(ui, repo, cost=1.0, func=b'runme'):
> cost = float(cost)
> func = functable[func]
> ui.status(b'start\n')
> runs = worker.worker(ui, cost, func, (ui,), range(8))
> for n, i in runs:
> pass
> ui.status(b'done\n')
> EOF
$ abspath=`pwd`/t.py
$ hg init
Run tests with worker enable by forcing a heigh cost
$ hg --config "extensions.t=$abspath" test 100000.0
start
run
run
run
run
run
run
run
run
done
Run tests without worker by forcing a low cost
$ hg --config "extensions.t=$abspath" test 0.0000001
start
run
run
run
run
run
run
run
run
done
#if no-windows
Known exception should be caught, but printed if --traceback is enabled
$ hg --config "extensions.t=$abspath" --config 'worker.numcpus=8' \
> test 100000.0 abort 2>&1
start
abort: known exception
[255]
$ hg --config "extensions.t=$abspath" --config 'worker.numcpus=8' \
> test 100000.0 abort --traceback 2>&1 | egrep '^(SystemExit|Abort)'
Abort: known exception
SystemExit: 255
Traceback must be printed for unknown exceptions
$ hg --config "extensions.t=$abspath" --config 'worker.numcpus=8' \
> test 100000.0 exc 2>&1 | grep '^Exception'
Exception: unknown exception
Workers should not do cleanups in all cases
$ cat > $TESTTMP/detectcleanup.py <<EOF
> from __future__ import absolute_import
> import atexit
> import os
> import time
> oldfork = os.fork
> count = 0
> parentpid = os.getpid()
> def delayedfork():
> global count
> count += 1
> pid = oldfork()
> # make it easier to test SIGTERM hitting other workers when they have
> # not set up error handling yet.
> if count > 1 and pid == 0:
> time.sleep(0.1)
> return pid
> os.fork = delayedfork
> def cleanup():
> if os.getpid() != parentpid:
> os.write(1, 'should never happen\n')
> atexit.register(cleanup)
> EOF
$ hg --config "extensions.t=$abspath" --config worker.numcpus=8 --config \
> "extensions.d=$TESTTMP/detectcleanup.py" test 100000 abort
start
abort: known exception
[255]
#endif