fix txb.sign having issues when an incomplete TX contains OP_0s, but not enough

[rubensayshi]

when I did my first PR to deal with the out of order signatures etc this was I think in there but we removed it because there was no use/test case that required it.

unfortunatly I had a client who was still using a old version of our SDK that used bitcoinjs-lib@1.4.4 and it produces only OP_0s for the precending pubkeys, so if you sign key with 2 - out of 3 - then you get OP_0 SIG_2 for signatures (and not OP_0 SIG_2 OP_0).

this fixes the signing breaking if this occurs.

testcase is hardcoded instead of fixtures because I had to start out with a partially signed half broken raw hex

you can cherry-pick blocktrail@442835b for 1.5.x version of this

[coveralls]

Coverage remained the same at 99.41% when pulling d2bee4d on blocktrail:sign-incorrect-op0s-master into f64df79 on bitcoinjs:master.

[dcousens]

Thanks for this @rubensayshi, I'll review this asap :)

[weilu]

@rubensayshi Would you consider move the signature check to ecdsa's verify function instead? It'd make the verify function more robust, and easier to test.

[dcousens]

@weilu hmmm, I can't decide if that is a good design or not? Right now I'd prefer to see an Error thrown (as it does) if the signature is missing?

[weilu]

@dcousens the rest of verify doesn't tell you why it's returning false, does it? :P And you know you hate "can't call foo on undefined" as much as I do. One solution I can think of is to pass an optional debug argument to verify which will populate the reason key when verify returns false.

[dcousens]

can't call foo on undefined

But, at least then I know the problem. I'll have a think on this one. I'm not overwhelmingly against it.

[dcousens]

ACK. Thanks @rubensayshi,