In https://cs.chromium.org/chromium/src/v8/test/fuzzer/regexp/ we already have some sample regexes. Do you suggest adding more?

Yes, it has ~1600 regexes whereas the current sample set is just 23. It should improve coverage hopefully.

How should the samples be provided? As individual files like currently, or is there some format?

There are two opportunities:

1) seed corpus: https://cs.chromium.org/chromium/src/testing/libfuzzer/fuzzers/BUILD.gn?q=regexp_fuzzer&sq=package:chromium&l=255

existing one is quite small: https://cs.chromium.org/chromium/src/v8/test/fuzzer/regexp/

but you can easily make it:
seed_corpus = [ "//v8/test/fuzzer/regexp/", "another_dir_with_testcaes", "and so on" ]

2) a dictionary (as Abhishek mentioned in the description)

existing one: https://cs.chromium.org/chromium/src/testing/libfuzzer/fuzzers/dicts/regexp.dict?q=regexp.dict&dr

documentation on the format: https://chromium.googlesource.com/chromium/src/testing/libfuzzer/+/HEAD/efficient_fuzzer.md#Fuzzer-Dictionary



The rule of thumb might be: Put relatively large inputs into seed corpus, use small constructions / expressions in dictionary.

Funny. The comment in the dictionary says I made it. I don't even remember.

I'll extract regexps from the test and add them as dict.