Long lasting fuzzing job: how to pause and keep current corpus?

[qathy]

Hello,

I currently have a fuzzing test running for quite a long time Fuzzing time: 118h19m43.680240575s and I was wondering:

• how can I obtain a refreshed status?
• is the status only output on the console when the corpus grows?
• is there a mean to stop the process while keeping the corpus found so far?
  Las time I stopped such a process, despite the option --corpus_database=./corpus/, it didn't write anything in my corpus directory.
  And at the time being, with Corpus size: 35, the same directory is still empty.
  Unfortunately, I'll soon need to reboot the host on which this process is running, and I really don't want to loose all the work done so far.

This binary was built after a git pull on Tuesday 19th May from fuzztest main branch.

--version output:

[.] Sanitizer coverage enabled. Counter map size: 76115, Cmp map size: 262144
fz_sec
Debug build (NDEBUG not #defined)

Any help from anyone would be warmly welcome.

Regards.

[qathy]

Still in the hope anyone has some hint on this aspect of fuzzing, I made some progress using the environment variable FUZZTEST_TESTSUITE_OUT_DIR (set to an absolute path) in place of the command-line option.

Doing so, I'm able to collect some interesting corpora.

With lots of environmental conditions, the data in this corpus may be

• minimized (combining both FUZZTEST_TESTSUITE_OUT_DIR and FUZZTEST_TESTSUITE_IN_DIR),
• extracted in a format usable as input for static tests (using only FUZZTEST_TESTSUITE_IN_DIR, and a dedicated and fake test).

But I was wondering if there's a standardized way to do those jobs without the need to switch from one environment to another.

I experienced the fact that depending on the architecture (ARM versus x64), either minimization or extraction works but never both, while the source code remains the same on both sides.