The addition of the pre-computed values for the ScalarBaseMult
optimizations added a benign race condition since a pointer to each
pre-computed Jacobian point was being used in the call to addJacobian
instead of a local stack copy.
This resulted in the code which normalizes the field values to check for
further optimization conditions such as equal Z values to race against the
IsZero checks when multiple goroutines were performing EC operations since
they were all operating on the same point in memory.
In practice this was benign since the value was being replaced with the
same thing and thus it was the same before and after the race, but it's
never a good idea to have races.
The writeElement function provides faster serialization for primitives.
This commit modifies all instances that call writeElement with a
pointer to a primitive or a byte slice to instead use the primitive /
writeVarBytes function so the faster serialization paths are used.
The benchmarks are still only compiled when running 'go test' so this has
no effect on regular usage. This is being done because benchmarks often
need access to internal state.
Normal tests are kept in a separate package since they also serve to
exercise the public API, and by intentionally making it more difficult to
reach into the internals, it helps ensure the public API is sanely usable.
Since the benchmarks can now access the internals directly, this commit
also removes the functions which exposed the internals to the test package
from internal_test.go which were only used by the benchmarks.
The comment for the RPC listen section in the sample-btcd.conf incorrectly
claimed that the default for the RPC server listener is to listen on all
interfaces by default. In reality, it only listens on localhost for IPv4
and IPv6 by default.
Closes#208.
The mempool's MaybeAcceptTransaction methods have also been modified
to return a slice of transaction hashes referenced by the transaction
inputs which are unknown (totally spent or never seen). While this is
currently used to include the first hash in a ProcessTransaction error
message if inserting orphans is not allowed, it may also be used in
the future to request orphan transactions from peers.
This commit adds a test for checking known good block scripts in a block
are valid. As a part of this, it adds the infastructure needed to load a
saved transaction store from a file which contains all of the input
transactions needed.
It also contains some changes from running goimports as well as some other
cleanup.
This commit modifies the getblocktemplate RPC to correctly handle the
curtime field.
Currently, the RPC server will refuse to serve a block template if the
current time is before the minimum allowed time for a block even though
the real generated block template already accounted for it.
This consists of three changes:
- Remove the unnecessary and incorrect check in the RPC invocation for the
current time against the min required time since it is already handled
properly by the block template generation.
- Expose the network time source to the RPC work state and use the
adjusted time for checking against the maximum allowed time instead of
the current time.
- Set the returned RPC result curtime field to the time of the generated
block header.
Fixes#209.
This commit modifies the creation of RPC results to ignore errors in
script disassembly since they already contain the error string inline and
the RPC results must still be generated regardless.
This was already done in the decodescript RPC, however the same thing was
not being done in the higher level verbose transaction results. This
applies to both the createrawtransaction and decoderawtransaction RPCs.
Also, since this was the only thing that could error within those
functions, the error returns and caller checking of now non-existent
errors have been removed.
Fixes#210.
The hash is not displayed big endian, but rather as a byte-reversed
hash as a hexidecimal string, so document it as such.
While here, speed up the function. Some benchmarks:
BenchmarkShaHashString 2000000 996 ns/op
BenchmarkShaHashStringOld 100000 22701 ns/op
This commit introduces a new flag, --notls, which can be used to disable
TLS for the RPC server. However, the flag can only be used when the RPC
server is bound to localhost interfaces. This is intended to prevent the
situation where someone decides they want to expose the RPC server to the
web for remote management/access, but forgot they have TLS disabled.
This commit corrects the error check from the createTxRawResult call in
handleGetRawTransaction. It was previously given a different name which
resulted in the wrong error being checked.
Fixes#196.
This commit improves a couple of issues surrounding the creation of the
btcd home directory.
First, the code was previously attempting to log any errors that occurred
while creating the directory using the logging system which is not
initialized at that point. Thus, nothing was displayed to the user.
Second, if any component of btcd home directory path already exists, but
is not a directory, such as in the case of symlinks, the error returned
from the os.MkDirAll call indicates the directory can't be created. While
this is true, it's not always the most helpful error to display to the
user. So, this commit adds logic to detect when the failure case is due
to an existing symlink and displays a nicer error message suggesting the
user check if the destination of the link is mounted.
Fixes#193.
This prevents a downgrade attack to the vulnerable SSLv3. While here,
go ahead and require at least TLS 1.2 since TLS 1.0 and 1.1 have their
own set of issues and it's only a matter of time before those would
need to be completely avoided as well.
ok @davecgh
Now that Go 1.4 has been released, this commit reverts the recent changes
which dealt with allowing TravisCI to work with go tip in between the Go
1.3 and Go 1.4 release cycle and updates the .travis.yml to invoke test
coverage tool from the new path in Go 1.4.
