PushedData returns an array of byte slices containing any pushed data
found in the passed script. This includes OP_0, but not OP_1 - OP_16.
help from and ok @owainga
This commit addes a couple of tests to ensure performing math operations
on a 5-byte integer, even if the result is a 4-byte integer, is considered
invalid.
Both the script tests (positive and negative) and tx texts (ditto) are
present. Some of the tx tests in the negative section have been
replaced by a comment line explaining why that test is elided, to add in
diffing. The reasons were always that they test things handled by other
parts of the btcd stack (normally chain). For example MAX_MONEY, number
of outputs, coinbase sizes etc.
Much of the inital test logic from @dajohi using hand transcribed tables
for selected tests. The json parsers, script format parser and a lot of
cleaning up/bugfixing from your truly. @davecgh had some input too.
This commit corrects the number of expected inputs for a multi-sig script
to include the additional item that is popped from the stack due to the
OP_CHECKMULTISIG consensus bug (which is required and properly performed).
Note this issue did NOT affect the consensus critical code and hence would
not cause a chain fork. It did however, cause standard p2sh multisig txns
to be rejected from the mempool as nonstandard.
The tx rejected as non-standard which prompted this was spotted by
@mbelshe on IRC.
ok @owainga
Found by tests dhill is working on. We checked that ifs were closed at the end
of execution but not at script switching time, we now move this to just after
finishing a single script.
This commit tightens the check for a pay-to-pubkey script by ensuring the
length of the pubkey is one of the two valid values of 33 or 65. This
mirrors the checks in the multisig script type check as well.
ok @owainga
Also, unexport the functions to generate script types. Everything should
(and is) be using PayToAddrScript() with an address type instead of
throwing bytes around.
discussed with #@davecgh
This commit adds a new ScriptBuilder interface that can be used to build
custom scripts. It currently is fairly basic, but it allows you to push
raw opcodes, ints, and data while respecting canonical encoding. These
primitives are sufficient to build any script.
This could be improved upon with quite a few things. One example would be
functions for certain opcodes that take properly typed parameters to make
it harder to create invalid scripts.
For now though, it is already quite useful since it handles all of the
opcode selection for canonical data pushes and integer encoding.
The initial discussion took place in #5.
This commit builds off the previous commit which fixed the execution of
multi-signature scripts with zero required signatures.
It introduces the concept of a "small int" which is one of OP_0 or OP_1 -
OP_16. All areas of code that deal with multi-sig transactions now make
use of these to ensure consistent handling.
This fixes a few issues surrounding multi-sig zero required signature
transactions included proper detection as a multi-sig script, signature
counting for script statistics, and
ok @owainga
It is possible for a multisignature transaction to require zero
signatures. For example, input 2 of testnet transaction
b2d93dfd0b2c1a380e55e76a8d9cb3075dec9f4474e9485be008c337fd62c1f7
in block number 185117.
Previously the code was pushing a false to the stack when no
valid signatures were found. This commit remedies that by pushing true
when no valid signatures were found, but none are required. Otherwise it
still pushes false when no valid signatures were found, but some are
required.
Fixes#7.
ok @owainga
When given badly formatted signature or pubkeys like in block
0000000000000001e4241fd0b3469a713f41c5682605451c05d3033288fb2244, transaction
fd9b541d23f6e9bddb34ede15c7684eeec36231118796b691ae525f95578acf1 we could fail
on strange scripts because we returned an error instead of failing the
opcode and putting a FALSE on the stack.
Fixes chainfork issue on the aforementioned block.
