Running through suspend/resume results in child timeout

[mpictor]

2020/07/01 01:47:16 workers: 16, corpus: 571 (8m47s ago), crashers: 0, restarts: 1/9966, execs: 64115646 (50405/sec), cover: 1118, uptime: 21m12s
2020/07/01 01:47:19 workers: 16, corpus: 571 (8m50s ago), crashers: 0, restarts: 1/9965, execs: 64269547 (50407/sec), cover: 1118, uptime: 21m15s
2020/07/01 18:26:06 workers: 16, corpus: 571 (8m54s ago), crashers: 0, restarts: 1/9967, execs: 64427609 (50359/sec), cover: 1118, uptime: 21m19s
2020/07/01 18:26:07 workers: 16, corpus: 571 (8m56s ago), crashers: 1, restarts: 1/9967, execs: 64589095 (50420/sec), cover: 1118, uptime: 21m21s
2020/07/01 18:26:10 workers: 16, corpus: 571 (8m59s ago), crashers: 1, restarts: 1/9965, execs: 64613976 (50322/sec), cover: 1118, uptime: 21m24s
2020/07/01 18:26:13 workers: 16, corpus: 571 (9m2s ago), crashers: 1, restarts: 1/9960, execs: 64760906 (50319/sec), cover: 1118, uptime: 21m27s

Note the time jump in the 3rd entry (when the machine was asleep) and the crasher reported in the 4th.

$ cat fuzz/crashers/1f5181fc9f7d217a15d78c5d439500aa64a5447f.output 
program hanged (timeout 10 seconds)

signal: killed

I renamed fuzz/crashers, copied the crasher input to the corpus, and restarted. This input does not cause crashes after the restart. I suspect there's a timer somewhere that doesn't compensate for lost time such as when the computer is asleep.

command used:

$ go-fuzz-build -func FuzzParseFilePathList github.com/u-root/u-root/pkg/uefivars/boot
$ go-fuzz -bin=./boot-fuzz.zip -workdir=fuzz

go-fuzz appears to be the latest version:

$ ls -l `which go-fuzz`
-rwxr-xr-x 1 mark mark 14M Apr 14 10:48 /home/mark/go/bin/go-fuzz
$ cd ~/go/src/github.com/dvyukov/go-fuzz
$ git log -n1 | cat
commit be3528f3a81351d8a438aed216130e1e7da39f7c
Author: Dmitry Vyukov <[email protected]>
Date:   Wed Mar 18 10:02:43 2020 +0100

    .travis.yml: fix osx setup
    
    There seems to be some change in osx machine setup on travis.
$ git pull
Already up to date.

[degeri]

Just confirming I am seeing the same. Suspend seems to throw the timeout calculation out of whack.

Tip: You can do this in a VM, replication will be easier.

example.mp4