You are adding around 20k, so I wonder if there is an overlap. Should I redo with your folder added?

Sure, feel free to pull it in, that’s what I pushed the branches for! :)

Ok, running that now.

Also, it looks like we are nearing the time limit here.

So, I'd say to merge this and then right after open the pull request to delete-nonreduced.sh?

Re-running CI: https://cirrus-ci.com/task/5867579709849600

Ok, running that now.

Also, it looks like we are nearing the time limit here.

So, I'd say to merge this and then right after open the pull request to delete-nonreduced.sh?

Approach ACK

Hmm, for some reason I am running out of disk space on my tmpfs /tmp. I'll run again over night and push tomorrow morning, if it works 🥲

Now the task is DONE, but it failed (:smiling_face_with_tear:)

#131072	pulse  cov: 3399 exec/s: 672 rss: 219Mb
#262144	pulse  cov: 6464 exec/s: 579 rss: 219Mb
#451694	DONE   cov: 7097 exec/s: 422 rss: 316Mb
MERGE-OUTER: successful in 1 attempt(s)
MERGE-OUTER: the control file has 4052306816 bytes
==1059184== ERROR: libFuzzer: out-of-memory (used: 2084Mb; limit: 2048Mb)
   To change the out-of-memory limit use -rss_limit_mb=<N>

SUMMARY: libFuzzer: out-of-memory

Ok, now it passed, with this patch applied to avoid out-of-memory and out-of-storage:

diff --git a/test/fuzz/test_runner.py b/test/fuzz/test_runner.py
index 9d64591111..591689b11c 100755
--- a/test/fuzz/test_runner.py
+++ b/test/fuzz/test_runner.py
@@ -278,6 +278,8 @@ def merge_inputs(*, fuzz_pool, corpus, test_list, src_dir, build_dir, merge_dirs
         args = [
             os.path.join(build_dir, 'src', 'test', 'fuzz', 'fuzz'),
             '-set_cover_merge=1',
+            '-rss_limit_mb=8000',
+            f'-merge_control_file=/run/media/marco/scratch_fde/fuzz/A_m_{t}.txt',
             # set_cover_merge is used instead of -merge=1 to reduce the overall
             # size of the qa-assets git repository a bit, but more importantly,
             # to cut the runtime to iterate over all fuzz inputs [0].

The first commits seems to add a bunch of files into the fuzz_seed_corpus directly instead of one of the subfolders. Any idea what may have happened there?

Can you give an example?

Both the first and the second commit’s I2P seeds pass on my machine:

That also seems funny. You could reproduce yesterday (at least in your folder), I can reproduce locally, and both CI tasks failed on your pull requests with this error.

Sorry, PEBCAK on the files. In my tig window, it dropped the fuzz target folder on some files in the change list, but not the others. I didn’t realize that it would present me different levels of the path in one list. 🤦
.

➜  qa-assets git:(main) ✗ git reset --hard maflcko/main                                   
HEAD is now at c764d6c7842 Revert i2p fuzz inputs to work around crash for now
➜  bitcoin git:(master) ✗ cd ../bitcoin/build                                            
➜  build git:(master) ✗ FUZZ=i2p src/test/fuzz/fuzz ../../qa-assets/fuzz_seed_corpus/i2p
i2p: succeeded against 1247 files in 27s.
➜  build git:(master) ✗ cd ../../qa-assets
➜  qa-assets git:(main) ✗ git reset --hard 00c956e4a9c555da5640946ec28cdd5e501c0554       
HEAD is now at 00c956e4a9c Add fuzz inputs from my servers, murch, and oss-fuzz
➜  qa-assets git:(main) ✗ cd -                                                            
~/Workspace/bitcoin/build
➜  build git:(master) ✗ FUZZ=i2p src/test/fuzz/fuzz ../../qa-assets/fuzz_seed_corpus/i2p
i2p: succeeded against 1256 files in 26s.
➜  build git:(master) ✗ 

I don’t know why it doesn’t crash on my machine but I can not replicate it crashing with either of your two commits here.

In any case, let's move discussion to bitcoin/bitcoin#28665

Which fuzzing engine are you using? Can you try with libFuzzer?

I've never seen this command and output:

FUZZ=i2p src/test/fuzz/fuzz ../../qa-assets/fuzz_seed_corpus/i2p
i2p: succeeded against 1247 files in 27s.

For reference, this increases the coverage from 63.6% (https://maflcko.github.io/b-c-cov/fuzz.coverage/index.html , unstable link) to 64.4% (https://drahtbot.space/host_reports/DrahtBot/reports/coverage_fuzz/monotree/091d29c49590d31d/c764d6c7842481a2/fuzz.coverage/index.html)

Could this PR cause CI timeouts?

See: bitcoin/bitcoin#28703.