Qemu signal refactoring

[rmalmain]

No description provided.

[tokatoka]

Can we change

impl Default for AsanModuleBuilder {
    fn default() -> Self {
        Self::new(
            None,
            false,
            true,
            StdAddressFilter::default(),
            None,
            AsanTargetCrash::OnFirstError,
        )
    }
}

to this?

impl Default for AsanModuleBuilder {
    fn default() -> Self {
    let env = std::env::vars()
        .filter(|(k, _v)| k != "LD_LIBRARY_PATH")
        .collect::<Vec<(String, String)>>();
        Self::new(
            None,
            false,
            true,
            StdAddressFilter::default(),
            None,
            AsanTargetCrash::OnFirstError,
        )
    }
}

else you unwrap this env later so it's easy to misuse

[rmalmain]

yes

[tokatoka]

maybe_report_crash calls run_observers_and_save_state.
and run_observers_and_save_state calls event_mgr.on_restart(state).unwrap();

it is mandatory to call on_restart to do restart but here you are only calling this in one of these branches. so you need to call in others too

[tokatoka]

I almost forgot what we discussed yesterday...
can you explain more on "In what case, you will end up in this branch?"

for example. if we receive a signal during harness execution, we go to which branch?

[tokatoka]

i guess this is the 1st signal?

[tokatoka]

and later InQemuSignalHandlerTarget is the 2nd signal after qemu delegated signal handling to us right?

[tokatoka]

Also can you talk about the asanmodules' api change in MIGRATION.md too?

[tokatoka]

i changed asan builder

[tokatoka]

seems not yet working

Fuzzer is working
rm -f sqltest static fuzz
+ ./tests/qasan/test.sh
gcc qasan.c -o qasan
Running overflow detection test...

and this job is stuck for the 217 minutes