Unable to use wally_tx_get_input_signature_hash from Java

[araspitzu]

Since release 1.4.0 I noticed the deprecation flag on several sighash calls and I started migrating to the new tx_get_input_signature_hash however I'm unable to make it work from Java via the JNI interface, every call results in a wally_map error. I've looked into the tests to see if I was doing something wrong and ended up recreating the taproot sighash test found in test_transaction.py in java and it still fails. I made two snippets for easy comparison of the test code java test and python test. Am I using the interface incorrectly? Note that I'm able to use Wally maps from java in another call tx_get_btc_taproot_signature_hash without issues.

[jgriffiths]

Hi @araspitzu thanks for the report and tests!

The issue is with the cache parameter; the java wrapper does not currently handle maps being optional and so throws when a null cache argument is given. You can make your test pass (and make signing multiple segwit/taproot inputs significantly faster) by adding:

Object cache = Wally.map_init(16);

And passing cache instead of null to the Wally.tx_get_input_signature_hash call.

I will add your test cases to the existing test suites and look at making the cache optional in the java case as it is for the other wrappers. Thanks again!

[jgriffiths]

Added your test and fixed cache optionality in #503

Thanks again for reporting!

[araspitzu]

Alright that fixed the issue! I didn't expect to have to pass a non-null cache but I guess it doesn't hurt to speed up hash generation for mulitple inputs. Thanks for the help @jgriffiths!