fix(test): remove lazy_static and call Drop on test exit

[luisschwab]

Closes #149.

This PR fixes the resource leakage issue on the test suite caused due to the BITCOIND and ELECTRSD statics not being Dropped on exit, and removes the lazy_static dev-dependency.

This is done by registering a cleanup routine via libc's atexit. Once the process terminates, it takes ownership of BITCOIND and ELECTRSD and calls drop() on them.

Verifying the fix:

1. Run the tests on master and check for any lingering bitcoind processes:

luisschwab@deepthought rust-esplora-client % just test
cargo test --all-features -- --test-threads=1
   Compiling esplora-client v0.12.1 (/Users/luisschwab/git/rust-esplora-client)
    Finished `test` profile [unoptimized + debuginfo] target(s) in 2.39s
     Running unittests src/lib.rs (target/debug/deps/esplora_client-49d1734ac197afec)

running 32 tests
test test::feerate_parsing ... ok
test test::test_get_address_stats ... ok
test test::test_get_address_txs ... ok
test test::test_get_address_utxos ... ok
test test::test_get_block_by_hash ... ok
test test::test_get_block_by_hash_not_existing ... ok
test test::test_get_block_hash ... ok
test test::test_get_block_info ... ok
test test::test_get_block_status ... ok
test test::test_get_block_txids ... ok
test test::test_get_block_txs ... ok
test test::test_get_blocks ... ok
test test::test_get_fee_estimates ... ok
test test::test_get_header_by_hash ... ok
test test::test_get_height ... ok
test test::test_get_merkle_block ... ok
test test::test_get_merkle_proof ... ok
test test::test_get_non_existing_block_status ... ok
test test::test_get_output_status ... ok
test test::test_get_scripthash_stats ... ok
test test::test_get_scripthash_utxos ... ok
test test::test_get_tip_hash ... ok
test test::test_get_tx ... ok
test test::test_get_tx_info ... ok
test test::test_get_tx_no_opt ... ok
test test::test_get_tx_outspends ... ok
test test::test_get_tx_status ... ok
test test::test_get_tx_with_http_header ... ok
test test::test_get_txid_at_block_index ... ok
test test::test_mempool_methods ... ok
test test::test_scripthash_txs ... ok
test test::test_that_errors_are_propagated ... ok

test result: ok. 32 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 15.50s

   Doc-tests esplora_client

running 2 tests
test src/lib.rs - (line 16) - compile ... ok
test src/lib.rs - (line 28) - compile ... ok

test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.05s

luisschwab@deepthought rust-esplora-client % ps aux | grep bitcoind
luisschwab       71177   0.0  0.0 410742832   1696 s000  S+    4:27PM   0:00.00 grep bitcoind
luisschwab       71161   0.0  0.2 411788208  62448 s000  S     4:27PM   0:00.70 /opt/homebrew/bin/bitcoind -datadir=/var/folders/7s/_67ys5jx23z_stpwj2nbz98r0000gn/T/.tmp3Zf4vi -rpcport=63607 -listen=0 -regtest -fallbackfee=0.0001

2. Kill the lingering bitcoind process and check no bitcoind process is left:

luisschwab@deepthought rust-esplora-client % killall bitcoind
luisschwab@deepthought rust-esplora-client % ps aux | grep bitcoind
luisschwab       71233   0.0  0.0 410733616   1584 s000  S+    4:29PM   0:00.00 grep bitcoind

3. Run the tests on this branch and check for any lingering bitcoind process:

luisschwab@deepthought rust-esplora-client % just test
cargo test --all-features -- --test-threads=1
   Compiling esplora-client v0.12.1 (/Users/luisschwab/git/rust-esplora-client)
    Finished `test` profile [unoptimized + debuginfo] target(s) in 0.68s
     Running unittests src/lib.rs (target/debug/deps/esplora_client-92c898aaf891e50d)

running 32 tests
test test::feerate_parsing ... ok
test test::test_get_address_stats ... ok
test test::test_get_address_txs ... ok
test test::test_get_address_utxos ... ok
test test::test_get_block_by_hash ... ok
test test::test_get_block_by_hash_not_existing ... ok
test test::test_get_block_hash ... ok
test test::test_get_block_info ... ok
test test::test_get_block_status ... ok
test test::test_get_block_txids ... ok
test test::test_get_block_txs ... ok
test test::test_get_blocks ... ok
test test::test_get_fee_estimates ... ok
test test::test_get_header_by_hash ... ok
test test::test_get_height ... ok
test test::test_get_merkle_block ... ok
test test::test_get_merkle_proof ... ok
test test::test_get_non_existing_block_status ... ok
test test::test_get_output_status ... ok
test test::test_get_scripthash_stats ... ok
test test::test_get_scripthash_utxos ... ok
test test::test_get_tip_hash ... ok
test test::test_get_tx ... ok
test test::test_get_tx_info ... ok
test test::test_get_tx_no_opt ... ok
test test::test_get_tx_outspends ... ok
test test::test_get_tx_status ... ok
test test::test_get_tx_with_http_header ... ok
test test::test_get_txid_at_block_index ... ok
test test::test_mempool_methods ... ok
test test::test_scripthash_txs ... ok
test test::test_that_errors_are_propagated ... ok

test result: ok. 32 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 15.26s

   Doc-tests esplora_client

running 2 tests
test src/lib.rs - (line 16) - compile ... ok
test src/lib.rs - (line 28) - compile ... ok

test result: ok. 2 passed; 0 failed; 0 ignored; 0 measured; 0 filtered out; finished in 0.06s

luisschwab@deepthought rust-esplora-client % ps aux | grep bitcoind
luisschwab       71389   0.0  0.0 410733616   1568 s000  S+    4:30PM   0:00.00 grep bitcoind

[coveralls]

Pull Request Test Coverage Report for Build 20795769652

Details

• 118 of 120 (98.33%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+0.3%) to 86.923%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
src/lib.rs	118	120	98.33%

Totals
Change from base Build 20725239190:	0.3%
Covered Lines:	1449
Relevant Lines:	1667

---

💛 - Coveralls

[ValuedMammal]

What I meant to suggest on the issue is to eliminate the use of static references and just initialize a new ElectrsD for every test (drop is already called when the object goes out of scope). At least that is the approach taken by the other BDK libraries, e.g. bdk_bitcoind_rpc, and keeps the tests from interfering with one another.

[luisschwab]

Oh right, let's just use bdk_testenv here as well. Much simpler.