You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Nov 1, 2023. It is now read-only.
Build a Linux sample libfuzzer (like simple-libfuzzer) with clang 15 (note: 14 should be sufficient), just invoking make as usual
Create a libfuzzer basic job with the target
Expected
Jobs run as expected, find crashes, and generate binary and source coverage.
Actual
The coverage task (only) fails unexpectedly with errors like:
[2022-04-11T22:02:34Z ERROR onefuzz_agent::managed::cmd] error running task: malformed debug info file
Caused by:
0: corrupted dwarf debug data
1: Hit the end of input before it was expected
Comments
Verified on Ubuntu 18.04, with targets built w/clang 15 from the bionic repo.
The crash analysis task succeeds in symbolizing stack traces. This appears to be an issue with our DWARF parsing libraries used for source coverage.
The default DWARF version emitted in clang when using -g (on Linux) was increased to v5 starting with LLVM 14, so this should also repro (not yet tested). The linked release notes note how to emit DWARF 4 as a workaround: -gdwarf-4 or -fdebug-default-version=4. This is verified to fix OneFuzz coverage analysis.
AB#35929
The text was updated successfully, but these errors were encountered:
Repro
simple-libfuzzer
) with clang 15 (note: 14 should be sufficient), just invokingmake
as usuallibfuzzer basic
job with the targetExpected
Jobs run as expected, find crashes, and generate binary and source coverage.
Actual
The coverage task (only) fails unexpectedly with errors like:
Comments
Verified on Ubuntu 18.04, with targets built w/clang 15 from the bionic repo.
The crash analysis task succeeds in symbolizing stack traces. This appears to be an issue with our DWARF parsing libraries used for source coverage.
The default DWARF version emitted in clang when using
-g
(on Linux) was increased to v5 starting with LLVM 14, so this should also repro (not yet tested). The linked release notes note how to emit DWARF 4 as a workaround:-gdwarf-4
or-fdebug-default-version=4
. This is verified to fix OneFuzz coverage analysis.AB#35929
The text was updated successfully, but these errors were encountered: