Dongge Liu
Dongge Liu
# Problem FuzzBench uses a cached fuzzer repo instead of the latest one, which sometimes does not include the new commits required. # Proof of Error https://github.com/google/fuzzbench/pull/1902#issuecomment-1737463393 # Proposed solution...
This was observed in [this experiment](https://github.com/google/fuzzbench/pull/1880#issuecomment-1681602328). Here is [the error log](https://pantheon.corp.google.com/logs/query;cursorTimestamp=2023-08-17T12:43:22.568776528Z;endTime=2023-08-21T01:38:11.678Z;query=jsonPayload.experiment%3D%222023-08-04-sfuzz%22%0AjsonPayload.benchmark%3D%22woff2_convert_woff2ttf_fuzzer%22%0Aseverity%3E%3DWARNING%0AjsonPayload.message!%3D%22Corpus%20not%20found%20for%20cycle:%200.%22%0Atimestamp%3D%222023-08-17T12:43:18.463572869Z%22%0AinsertId%3D%2218yfmbdffhas18%22;startTime=2023-08-16T01:38:11.678Z?project=fuzzbench): ``` Traceback (most recent call last): File "/src/experiment/runner.py", line 454, in experiment_main runner.conduct_trial() File "/src/experiment/runner.py", line 276, in conduct_trial...
Recent two new experiments in `service/experiment-requests.yaml` were not automatically triggered: 1. #1852 2. #1855 They can be launched via `GCB` commands: 1. #1853 2. #1856
(Not need to merge: This PR is for experiment purpose only) # Coverage ## Fuzzers 1. `hastefuzz` 2. `aflplusplus` 3. `aflplusplus_d82218` (new version) 4. `aflplusplusplus` 5. `aflplusplusplus_d82218` (based on new...
Add a metric to measure the time taken to find the bug in bug-based benchmarks.
1. Suppress a future warning on `seaborn.lineplot`'s `ci`. 2. Suppress a deprecation warning on `Orange`. 3. Adjust a warning on invalid benchmarks.
1. Deprecate [the list of fuzzers as discussed](https://github.com/google/fuzzbench/pull/1526#issuecomment-1334641934). 2. Let CI [ignore fuzzers](https://github.com/google/fuzzbench/pull/1570/files#diff-c0fba10e89a75fc5b00739b8d8a7866a703c0c37032f2c4a415ba1027e6242aeR113) that we want to keep but are not yet compatible with Ubuntu20 or Python3.10.
The API returns an empty list of functions when a project failed to build with FI, for example: ``` # https://introspector.oss-fuzz.com/api/far-reach-but-low-coverage?project=arrow {"functions":[],"result":"success"} ``` However, it's hard to tell the why...
The return type of function [`std::string CEscape(absl::string_view src);`](https://github.com/abseil/abseil-cpp/blob/8ebad34c3fa54a9ad2f46ca8cab98e75c4f750bf/absl/strings/escaping.h#L90) was incorrect in [the report.](https://storage.googleapis.com/oss-fuzz-introspector/abseil-cpp/inspector-report/20230823/summary.json) It should be `std::string`, but the report records `void`.
19 instances from [the same experiment](https://llm-exp.oss-fuzz.com/Result-reports/scheduled/2024-03-02-weekly-all/sample/output-kamailio-main2/01) do not terminate in days. Their symptoms indicate different causes. * Fail to capture a terminated instance. In [this example](https://llm-exp.oss-fuzz.com/Result-reports/scheduled/2024-03-02-weekly-all/benchmark/output-kamailio-handle_tcpconn_ev), the status of the...