Only libFuzzer is supported right now, so clang is required. Thus, this needs a workaround such as #12602 (comment)
Can be tested with:
mkdir build && cd build
../configure --enable-fuzz --with-sanitizers=fuzzer --enable-lcov --enable-lcov-branch-coverage CC=clang CXX=clang++
make $MAKEJOBS
make cov_fuzz
Strongest possible Concept ACK -- will test :)
BTW, I noticed that the fuzzing coverage link is broken on your otherwise excellent coverage statistics page. Will the merge of this fix that? :)
Which link is broken? https://marcofalke.github.io/btc_cov/ ?
Note that the fuzzers were running into a stack overflow, thus unable to produce a result. So for a while that page was a 404. This should now be fixed with 657c5e5f1cd92169b5a3c880f01376b7ed0fd9b3
@MarcoFalke Ah, then the issue was temporary. Cannot reproduce today :)
Strongest possible Concept ACK -- will test :) @practicalswift Are you still planning to test? :thinking:
@MarcoFalke Absolutely :)
I ran in to some problems - see below.
What versions of gcov, lcov and clang++ are you using? I'm using gcov 7.5.0, lcov 1.13 and clang++ 10.0.0.
$ make distclean
$ ./autogen.sh
$ ./configure --enable-fuzz --with-sanitizers=fuzzer --enable-lcov --enable-lcov-branch-coverage CC=clang CXX=clang++
$ make
$ git clone https://github.com/bitcoin-core/qa-assets
$ make cov_fuzz
…
Run uint256_deserialize with args ['…/src/test/fuzz/uint256_deserialize', '-runs=1', 'qa-assets/fuzz_seed_corpus/uint256_deserialize']
Output: INFO: Seed: 1103855062
INFO: Loaded 1 modules (1032 inline 8-bit counters): 1032 [0x5624c7ccdb08, 0x5624c7ccdf10),
INFO: Loaded 1 PC tables (1032 PCs): 1032 [0x5624c7ccdf10,0x5624c7cd1f90),
INFO: 0 files found in qa-assets/fuzz_seed_corpus/uint256_deserialize
INFO: -max_len is not provided; libFuzzer will not generate inputs larger than 4096 bytes
INFO: A corpus is not provided, starting from an empty corpus
[#2](/bitcoin-bitcoin/2/) INITED cov: 119 ft: 120 corp: 1/1b exec/s: 0 rss: 27Mb
[#2](/bitcoin-bitcoin/2/) DONE cov: 119 ft: 120 corp: 1/1b lim: 4 exec/s: 0 rss: 27Mb
Done 2 runs in 0 second(s)
/usr/bin/lcov --gcov-tool=/usr/bin/gcov -c --rc lcov_branch_coverage=1 -d …/src --t fuzz-tests -o fuzz.info
Capturing coverage data from …/src
Found gcov version: 7.5.0
geninfo: WARNING: invalid characters removed from testname!
Scanning …/src for .gcda files ...
Found 361 data files in …/src
Processing src/libbitcoin_util_a-fs.gcda
…/src/libbitcoin_util_a-fs.gcno:version '402*', prefer 'A75*'
geninfo: ERROR: GCOV failed for …/src/libbitcoin_util_a-fs.gcda!
Makefile:1321: recipe for target 'fuzz.info' failed
make: *** [fuzz.info] Error 29
rm baseline_filtered.info fuzz.info baseline.info
$ lcov --version
lcov: LCOV version 1.13
$ gcov --version | head -1
gcov (Ubuntu 7.5.0-3ubuntu1~18.04) 7.5.0
$ clang++ --version | head -1
clang version 10.0.0 (https://github.com/llvm/llvm-project.git f1b117394d7f9ae6decf9730ed9d443ca1b54769)
@practicalswift Have you applied the workaround? #12602 (comment)
Other than that I use vanilla debian:bullseye with whatever clang comes in there.
@MarcoFalke Thanks! Works as expected with workaround applied.
ACK faf7d4fa86b700ec272806cd2bd8666a92405619