Add fuzz seeds from 25rc1

[murchandamus]

Draft: Testing whether I’m doing this right so far, before I throw more compute at fuzzing.

[maflcko]

Usually I use current running master. I don't think there was ever an instance where an input on a release branch triggered a bug that wouldn't trigger on master as well, and vice-versa. Though, if there are meaningful backports to a release branch, this could change.

So Concept ACK. I presume you merged this with 25rc1 compiled and not master?

[murchandamus]

Yes, exactly, I was fuzzing with 25rc1. I’ve been fuzzing more while I was traveling, only to learn that libfuzzer appears to take the first of either -runs or -max_total_time and you can’t combine the two to make it stop whenever you hit one of the two limits. So… was fuzzing one thing for 35h, instead of everything for up to 20 minutes. Will merge my results on Monday and update this PR then.

[murchandamus]

Fuzzed every target for another 20 minutes.

[maflcko]

about wallet_notifications: #121 (comment)

[murchandamus]

@MarcoFalke: Damn, that was one of the things that were really slow to fuzz. ;)
So, do I understand you right that you are asking that I delete the content of the wallet_notifications target?

[maflcko]

Yeah, for now. Longer term it might be good to rework the fuzz target or come up with a different solution.

[murchandamus]

Updated to remove wallet_notifications fuzz seeds

[maflcko]

this: https://drahtbot.space/host_reports/DrahtBot/reports/coverage_fuzz/monotree/fa95970671c2dc22/f724b921fcddfaac/fuzz.coverage/index.html

nightly: https://marcofalke.github.io/b-c-cov/fuzz.coverage/index.html

[maflcko]

Can't really find new coverage, but I guess it doesn't hurt too much either?

lgtm

[fanquake]

@dergoegge

[dergoegge]

lgtm

[murchandamus]

Can't really find new coverage, but I guess it doesn't hurt too much either?

lgtm

I’m confused. If these new files didn’t increase the coverage, why did the merge not remove them? Did I do it wrong? :-/

[maflcko]

The merge uses -use_value_profile=1, IIRC, and the fuzz engine may use some other stuff internally, which triggers on more than just Function/Line/Branch coverage and may include fuzz inputs that appear otherwise "redundant".