Negative int32 values are not decoded properly.

[duaneellissd]

Module: encodings.js - function int32 - is wrong.
It does not properly decode negative numbers.

For example "-15" encodes as 10 bytes
0xf1 0xff 0xff 0xff 0xff 0xff 0xff 0xff 0xff 0x01

The return value from "varint" is a giant number
As each byte is decode, the result value becomes:
BYTE is: 241 res: 0
BYTE is: 255 res: 113
BYTE is: 255 res: 16369
BYTE is: 255 res: 2097137
BYTE is: 255 res: 268435441
BYTE is: 255 res: 34359738353
BYTE is: 255 res: 4398046511089
BYTE is: 255 res: 562949953421297
BYTE is: 255 res: 72057594037927920
BYTE is: 1 res: 9223372036854776000
Result: 18446744073709552000
int32, value = 18446744073709552000 (Which is the unsigned value)

Above is the proper return value from "varint" the problem is in the encoding.js file

The 'pseudo-negative' test is correct, ie: if the value is > 2147483647 [then the value is negative]

however , the code that should adjust for a negative number is just wrong.

This is with "node.exe" windows, v4.2.6, on Win7 64bit, I have not tried other platforms.

[mafintosh]

Can you share a full runnable example that reproduces your above issue? If I encode -15 using the int32 function i get 5 bytes, [241, 255, 255, 255, 15] which, when decoded using int32, yields -15

[duaneellissd]

Can't easily give you a 'working example', two reasons (1) the example comes from "protobuf-c" - a C implementation that talks to the NodeJS solution.

using the int32 function i get 5 bytes

Reason (2) The encoding scheme in your implementation is wrong, this should encode as 10 bytes. See the protocol buffers specification here: https://developers.google.com/protocol-buffers/docs/encoding#structure

The quote in the specification you are looking for is:

If you use int32 or int64 as the type for a negative number, the resulting varint is always ten bytes long – it is, effectively, treated like a very large unsigned integer.

This is the basic cause of the inefficient encoding the other pages refer to.

[duaneellissd]

I presume the reference to the specification (i.e.: negative numbers encode as 10 bytes, not 5) its sufficient to explain the problem?

Do you need anything else?

Thanks.

[mafintosh]

Yep I understand it. You wanna take a stab at fixing this?