Copy secp256k1 code from go-ethereum to avoid GPL vendoring issues in

[silasdavis]

This copies the CGO secp256k1 implementation from https://github.com/ethereum/go-ethereum/tree/729bf365b5f17325be9107b63b233da54100eec6/crypto/secp256k1 added as dependency by #3234.

The code in question is MIT not GPL licensed but dep will still pull in the entire go-ethereum library. This causes and issue for downstream projects like http://github.com/hyperledger/burrow that vendor Tendermint and are permissively licensed. Pruning would sort-of solve this but doesn't work for CGO code, and in any case for code that is as stable over time as this is I think a little duplication is cleaner.

From my conversation with @liamsi I believe this was plan anyway, so here we are :)

Some other justifications for this:

• go-ethereum doesn't own the core code in this package and they haven't update it for 2 years, so there is little to gain from having it as a formal dependency
• copying the bit that is needed in means that you won't unncessarily create conflicts with any downstream projects that would like to depend on a version of go-ethereum of their choosing (cough ethermint cough)
• Managing go packages doesn't get any more painful for me in Burrow :)

[codecov-io]

Codecov Report

Merging #3371 into develop will increase coverage by 0.18%.
The diff coverage is n/a.

@@             Coverage Diff             @@
##           develop    #3371      +/-   ##
===========================================
+ Coverage    64.23%   64.41%   +0.18%     
===========================================
  Files          222      214       -8     
  Lines        17828    17416     -412     
===========================================
- Hits         11451    11219     -232     
+ Misses        5435     5266     -169     
+ Partials       942      931      -11

Impacted Files	Coverage Δ
consensus/reactor.go	70.95% <0%> (-0.83%)	⬇️
consensus/state.go	78.6% <0%> (-0.24%)	⬇️
libs/db/remotedb/grpcdb/server.go	0% <0%> (ø)	⬆️
lite/proxy/proof.go
rpc/lib/types/types.go
crypto/secp256k1/secp256k1_nocgo.go
rpc/client/mock/client.go
lite/proxy/wrapper.go
crypto/secp256k1/secp256k1.go
lite/proxy/errors.go
... and 12 more

[liamsi]

ref: #3289

[liamsi]

I've verified that the code under crypto/secp256k1/internal/secp256k1 in this PR is basically the same as in https://github.com/ethereum/go-ethereum/tree/729bf365b5f17325be9107b63b233da54100eec6/crypto/secp256k1
"Basically" because there is a test-file which was deleted secp256_test.go and comment added to the readme:

+++ b/crypto/secp256k1/internal/secp256k1/README.md
@@ -0,0 +1,3 @@
+This package is copied from https://github.com/ethereum/go-ethereum/tree/729bf365b5f17325be9107b63b233da54100eec6/crypto/secp256k1
+
+Unlike the rest of go-ethereum it is MIT licensed so compatible with our Apache2.0 license. We opt to copy in here rather than depend on go-ethereum to avoid issues with vendoring of the GPL parts
 of that repository by downstream.

https://github.com/tendermint/tendermint/blob/5f93d5392ddb71009c43fad3fc07d0fa49d2cc66/crypto/secp256k1/internal/secp256k1/README.md

Codeowners (@melekes, @ebuchman) can verify this themselves by doing the following:

1. checkout the PR changes locally (e.g. via git fetch origin pull/3371/head:silasdavis/internalise-libsecp256k1 && git checkout silasdavis/internalise-libsecp256k1)
2. temporarily get go-ethereum repo: go get -v -d github.com/ethereum/go-ethereum
3. cd $GOPATH/src/github.com/ethereum/go-ethereum/ and checkout the commit Silas' mentioned: git checkout 729bf365b5f17325be9107b63b233da54100eec6 (will be in detached head, cleanup or delete afterwards)
4. go back to tendermint cd $GOPATH/src/github.com/tendermint/tendermint/
5. git diff $GOPATH/src/github.com/ethereum/go-ethereum/crypto/secp256k1/ crypto/secp256k1/internal/

Here is the full output of above operation (created via git diff --no-prefix $GOPATH/src/github.com/ethereum/go-ethereum/crypto/secp256k1/ crypto/secp256k1/internal/ > diff_from_ethereum.patch):

diff_from_ethereum.txt

[liamsi]

See: #3371 (comment)

[ebuchman]

Verified.

But why was the test file deleted?