"python ./runtest.py" is broken about pyethereum HOT 29 CLOSED

it appears it's a json parsing problem. treat it as a dictionary?

from pyethereum.

yeah .. looking into it now

from pyethereum.

off-topic question: is there an an irc channel where ethereum devs hang out?

from pyethereum.

not that i know of, try #ethereum or #ethereum-dev

from pyethereum.

Here's what I get on OSX:

pyethereum $ python ./runtest.py
Traceback (most recent call last):
  File "./runtest.py", line 7, in <module>
    rlpdata = json.loads(open(os.path.join(testdir,'rlptest.txt')).read())
IOError: [Errno 2] No such file or directory: '../tests/rlptest.txt'

I can't quickly find out where this rlptest.txt is supposed to come from.

(and yes both #ethereum and #ethereum-dev is active)

from pyethereum.

You need to clone the "tests" which can be found here: https://github.com/ethereum/tests

from pyethereum.

OSX Mavericks is throwing this error:

$ pwd
$HOME/pyethereum/tests       // I cloned ethereum/tests into tests
$ python ../runtest.py
Mismatch with adds only ("{u'dogglesworth': u'cat', u'dog': u'puppy', u'doe': u'reindeer'}" -> "cf7d318935b52db6e23d8c1f5e6b7e62f3606d4ed13783f4fdbd6e67a2085d04")
Mismatch with adds only ("{u'do': u'verb', u'horse': u'stallion', u'doge': u'coin', u'dog': u'puppy'}" -> "0d9ce42de1d9e290181283ff8168f9dc40b93de6d8a06f55670290881cc95f2d")

from pyethereum.

That's expected -- the runtest.py script was fixed but the test data in the "tests" repository wasn't updated.

Please note also that a large part of runtest.py was ported to "proper" python tests (see the "tests" subdirectory). These use the same test data and will not pass either yet.

In a next step I'd like to drop the dependency on the external "tests" repository and have the test data reside local to "pyethereum".

from pyethereum.

Could someone help me out here with all the tests involved?

@freizeit in #11 you introduce py.test and you reference this (#7) as fixed. However it is still open and runtest.py is still in the repo. Furthermore in #12 you reference #7 as fixed.

@chenhouwu In commit 4703540 you introduced 'behave' as new test framework, but runtest.py is still in the repository.

If I understand correctly, py.test and runtest.py still rely on cloning the tests repo (and fail if not cloned)?

So...,

• is there any clear roadmap towards a single test framework?
• when creating PR's, which tests do my commits need to pass?

Thanks!

from pyethereum.

Currently only test for rlp is implemented in behave,
so run behave will only test rlp.

trie and others are still same as before, but it surely will be migrated to behave later.
so currently runtest.py shoulde be modified now to get rid of the codes for rlp testing

from pyethereum.

runtest.py is obsolete and removed

from pyethereum.

Hello there! Sorry I could not answer all the questions in time. My understanding is that the rlp module is broken and does not comply to https://github.com/ethereum/wiki/wiki/%5BEnglish%5D-RLP

I am in the process of preparing a branch that will address some of this.

Could you all please also join https://groups.google.com/forum/#!forum/ethereum-dev so we can discuss and agree on development issues? Thank you very much indeed!

from pyethereum.

@konradkonrad yes, the py.test tests are still cloning the "tests" repository. In the "rlp fixes" branch (currently in progress) I test against the "develop" branch.

However, I have doubts re. the following values in https://github.com/ethereum/tests/blob/develop/rlptest.json

• "zero"
• "bigint"

and will prepare an email to the ethereum-dev to discuss this.

Regarding test frameworks: I have used py.test with good success in the past and know the author well -- hence my inclination to use it for pyethereum as well.

from pyethereum.

I've joined the google groups :-).

rlp encoding would only encode string (ie. byte array) and list.
So an integer first should be converted to a string.
Thus the encoding result depends on how it's converted.

If '{0}'.format(zero) then the result will be \x30,
but if chr(zero), the result would be \x00

from pyethereum.

In the Ethereum(++) client \x00 is encoded as empty string "" when sending rlp encoded data over the network. Therefore "zero" and "emptystring" have the same value "80".

from pyethereum.

Here's how utils.py should look like when encoding "zero" as the empty string.
https://gist.github.com/heikoheiko/9727917

from pyethereum.

@heikoheiko I understand now -- thanks for the clarification !

from pyethereum.

Another question: encoding as follows is clear:

int -> byte buffer -> RLP

However, the other way around

RLP -> byte buffer

How would I know whether the byte buffer should further be transformed into an integer or left as is?

from pyethereum.

Further transformation depends on the application. You basically have to know what data and types you expect to get out of the deserialization.

For example if you get ['Bob', '7', ''] from rlp.decode, and know that you expect (str:name, int:account, int:balance) then you would have to treat the last two elements as big-endian encoded integers and finally get ['Bob', 55, 0].

from pyethereum.

@al-maisan , sorry, just mix the form of '%d' % (0) and '{0}'.format(0) in the previous post.
Already fixed now

@heikoheiko, the utils.py fike is created by me,

de5d6c98 utils.py (Chen Houwu 2014-03-18 12:57:42 +0800  1) def int_to_big_endian(integer):
de5d6c98 utils.py (Chen Houwu 2014-03-18 12:57:42 +0800  2)     '''convert a integer to big endian binary string
de5d6c98 utils.py (Chen Houwu 2014-03-18 12:57:42 +0800  3)     '''
de5d6c98 utils.py (Chen Houwu 2014-03-18 12:57:42 +0800  4)     s = '%x' % integer
f2b796b9 utils.py (Chen Houwu 2014-03-17 16:04:36 +0800  5)     if len(s) & 1:
f2b796b9 utils.py (Chen Houwu 2014-03-17 16:04:36 +0800  6)         s = '0' + s
f2b796b9 utils.py (Chen Houwu 2014-03-17 16:04:36 +0800  7)     return s.decode('hex')
de5d6c98 utils.py (Chen Houwu 2014-03-18 12:57:42 +0800  8)
de5d6c98 utils.py (Chen Houwu 2014-03-18 12:57:42 +0800  9) def big_endian_to_int(string):
de5d6c98 utils.py (Chen Houwu 2014-03-18 12:57:42 +0800 10)     '''convert a big endian binary string to integer
de5d6c98 utils.py (Chen Houwu 2014-03-18 12:57:42 +0800 11)     '''
de5d6c98 utils.py (Chen Houwu 2014-03-18 12:57:42 +0800 12)     s = string.encode('hex')
de5d6c98 utils.py (Chen Houwu 2014-03-18 12:57:42 +0800 13)     return int(s, 16)

I don't give the special case when the value is 0.
So the following codes are not there

if integer == 0:
  return '' 

from pyethereum.

@al-maisan for RLP encoding, it only accept string(i.e raw byte) or a list of string. RLP itself is designed to work in low level without the know what the strings means. So how the high level data is serialized to string and deserialized from string is totally up to you.

Hope it's clarified now.

from pyethereum.

@chenhouwu in order to be compatible with the Ethereum(++) client (which is the most current spec) we need to treat the special case of zero/empty_string. i.e. big_endian_to_int("") will fail. I fixed this here: https://github.com/brainbot-com/pyethereum/blob/d2e166a6b8196e0cc24bf6f2fe2dcaab3a578506/pyethereum/utils.py

from pyethereum.

@heikoheiko yes, blank string will be encoded to \0x80. but for zero, first we should convert it into byte array. so here you mean zero is converted to ''. but 1 or 2 or some other integers are not.
That's inconsistent obviously.

from pyethereum.

in fact, in the RLP encoding spec, it does not care about how you convert integer in high level codes.
For integer, it only takes care of the result of "payload length plus some prefix(like 0x80, 0xc0...)"

So in the low level RLP encoding, it will never meet the case of convert 0 to big endian array.

from pyethereum.

@al-maisan

If '{0}'.format(zero) then the result will be \x30(not \x80),
but if chr(zero), the result would be \x00.

So, why the \x80 comes up, I don't know.
I'll discuss it with @vbuterin further

from pyethereum.

@chenhouwu can you please take that discussion to https://groups.google.com/forum/#!forum/ethereum-dev so other people see it as well and we document the outcome and the discussion? Thanks!

P.S.: @vbuterin has subscribed to that list as well

from pyethereum.

The way things should be:

rlp.encode(0) = rlp.encode("") = '\x80'
rlp.encode(chr(0)) = rlp.encode("\x00") = '\x00'
rlp.encode([]) = '\xc0'

There is no need to special case zero; it's just a natural extension of the rules for nonzero length strings. For integer conversion, 0 -> '' is just the base case.

from pyethereum.

@vbuterin
Still not clear about the conversion, why rlp.encode(0) = rlp.encode("")
i.e. in what rule integer 0 is converted to ""?
also, what's the result of rlp.encode(10)?

from pyethereum.

From @vbuterin

rlp.encode(x) = chr(x) holds for 1 <= x <= 255

and rlp.encode(0) = rlp.encode("") is a pre-determined case.

from pyethereum.