the cold-outputs and cold-transaction fuzz tests in monero cancel the AFL persistent-mode loop if an import or sign, respectively, fails

since this happens pretty often with randomly-generated inputs, that means that the persistent-mode loop is basically not used

and thus that you have <1 execution per second, which makes fuzzing only questionably useful

do I understand this correctly / is this known?

I don't think anyone runs those. I don't recall anything running at anywhere close to 1/s, though I think when I added persistent mode I did not have support for it.

Feel free to fix.

initialisation of the wallet takes ~1.5 seconds on the machine I was trying it on

Oh, that's probably the subaddresses init. See unit_tests for how to punch it in the face.

(the wallet initialisation slowness is also an issue for the 'signature' fuzz test, but there at least it properly loops with persistent mode so it isn't as much of a performance problem)

wallet.set_subaddress_lookahead(1, 1);

cold-outputs and cold-transaction seem to call that

Then maybe the SSL cert.

Hmm. Also handles that. I dunno then.

Oh. Randomx

Might be the dataset init.

running under perf doesn't give me anything useful but that's probably a lack of symbols

wait does the 'fuzz' target not compile in release mode?

I'll be running through merges again tommrow. Main PC decided to brick itself yesterday, so couldn't continue. I've got my massive list-o-things to merge though.

Thanks

This: loki-project/loki 6cf5668 adds a full second of delay on start up