The biggest change here is that we're going from .GCDA files to .PROFRAW during coverage generation (see #110).

@R9295 can you please test this out on a few projects and tell me if everything works as expected?

@vanhauser-thc this PR reverts the dual minimization PR and fixes all of the duplicate seed issues:

1. After any minimization, a name-hashing pass is done over all of the minimized corpus.
2. During fuzzing, honggfuzz and afl corpora are copied-over to the shared corpus every 10 seconds, and the file names are the hash of the file itself.

Could you please take a quick look?

The biggest change here is that we're going from .GCDA files to .PROFRAW during coverage generation (see #110).

@R9295 can you please test this out on a few projects and tell me if everything works as expected?

I evaluated this too. the coverage I obtained was quite different between the two. have you looked into that?

done with my first review :)

The biggest change here is that we're going from .GCDA files to .PROFRAW during coverage generation (see #110).
@R9295 can you please test this out on a few projects and tell me if everything works as expected?

I evaluated this too. the coverage I obtained was quite different between the two. have you looked into that?

coverages.zip

Here are the differences I could see at first glance:

1. The new profraw method sees more dependencies (11 vs 8 in this url example), I have not figured out why
2. profraw sees some super interesting stuff that gcda does not (see url-2.5.2/src/parser.rs)
3. it seems the two count coverage for a multi-line expression differently (see tinyvec-1.8.0/src/arrayvec.rs)

Doesn't look like a downgrade, and kind of looks like an upgrade in some places.

the changes even use less lines of code :) good job!