Make fuzz runtime seconds not iterations #4179
Conversation
|
I've assigned @wpaulino as a reviewer! |
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## main #4179 +/- ##
=======================================
Coverage 88.83% 88.83%
=======================================
Files 180 180
Lines 137504 137504
Branches 137504 137504
=======================================
+ Hits 122155 122156 +1
+ Misses 12538 12536 -2
- Partials 2811 2812 +1
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. 🚀 New features to boost your workflow:
|
TheBlueMatt
force-pushed
the
2025-10-fiz-fuzz
branch
from
cbb53b6 to
3a4387d
Compare
TheBlueMatt
force-pushed
the
2025-10-fiz-fuzz
branch
5 times, most recently
from
8bfffb2 to
1ba5def
Compare
| # Because we're fuzzing relatively few iterations, the maximum possible | ||
| # compiler optimizations aren't necessary, so we turn off LTO | ||
| sed -i 's/lto = true//' Cargo.toml | ||
| sed -i 's/codegen-units = 1//' Cargo.toml |
There was a problem hiding this comment.
In #3916 we made some attempts of benchmarking before we removed this line. Do you have a vague number of how much the slowdown would be?
There was a problem hiding this comment.
I forgot about that, but its almost certainly a tiny difference in fuzzing performance, for a bit more parallelism while compiling some crates. 🤷♂️
TheBlueMatt
force-pushed
the
2025-10-fiz-fuzz
branch
from
1ba5def to
89671a5
Compare
|
Oh, weird, I guess bumping the debian version fixed the build even with 1.75, which I thought I tested....anyway, I dropped the MSRV bump for fuzzing cause it appears to be woring now. Now this just reduces the runtime to make it more consistent across jobs. |
TheBlueMatt
changed the title
It's failing in CI with a linking error though? |
We have some complexity in `ci-fuzz.sh` to limit each fuzzer to a rough runtime, but `honggfuzz` has a `--run-time` argument that we can simply use instead, which we do here.
This now slows us down as we run our fuzz job on a machine with more than one or two cores.
TheBlueMatt
force-pushed
the
2025-10-fiz-fuzz
branch
from
89671a5 to
fbdb5c5
Compare
|
Weird, it seems to be inconsistent, sometimes its fine, sometimes it fails. |
3 participants
No description provided.