Filter coverage recording against human-readable, demangled symbols.
- Add custom demanglers for Itanium C++ mangling, rustc mangling, and MSVC decorated names
- Add a catch-all demangler that tries each known demangler against a raw symbol, in a fixed order
- Default to using the catch-all demangler in coverage recording
We try to implement a lowest common denominator across schemes: omit types and extra annotations, but preserve generic specializations, namespacing, and paths. Note that the omission of parameter types causes collisions in the face of ad hoc polymorphism. Consult the unit tests for examples.
Update the filter rule format and implementation to be simpler and user-predictable. In particular, we remove an accidental dependence of rule application on hash map iteration order.
- Add caching to symbol table-driven module disassembly on Linux.
- Add configurable regex-based filtering for coverage collection, by module and module-scoped symbol name.
Block coverage recording can be manually tested using the `block_coverage` example in the `coverage` crate. See `./block_coverage -h` for expected args.
The filter file is optional. The file format is JSON like this:
```json
{
"modules": {
"allow": [
"<module-path-regex-1>",
"<module-path-regex-2>",
]
},
"symbols": {
"<module-path-regex-1>": {
"allow": [
"<symbol-name-regex-1>",
"<symbol-name-regex-2>",
]
},
"<module-path-regex-2>": {
"deny": [
"<symbol-name-regex-3>",
"<symbol-name-regex-4>",
]
}
}
}
```
Closes#285.
In debugging the connection retry issues, I dug into this more.
Apparently, some of hyper's connection errors are not mapped to std::io::Error, rendering the existing downcast impl ineffective.
As such, this PR makes the following updates:
1. Any request that fails for what `reqwest` calls a `connection` error is considered transient.
2. Updates the retry notify code to use our `warn` macro such that the events show up in application insights.
3. Updates the unit test to demonstrate that failures by trying to connect to `http://localhost:81/`, which shouldn't be listening on any system.
4. Adds a simple unit test to verify with send_retry_default, connections to https://www.microsoft.com work
Fixes#263
* Documents `crashes_account` and `crashes_container`
* Adds `reports_dir` and support for `unique_reports`, `reports`, and `no_repro` containers to the generic analysis task
* Adds `microsoft_telemetry_key` and `instance_telemetry_key` to generic supervisor, generator, and analysis tasks
In the supervisor agent, incrementally read from the running worker agent's redirected stderr and stdout, instead of waiting until it exits.
The worker agent's stderr and stdout are piped to the supervisor when tasks are run. The supervisor's `WorkerRunner` does _not_ use `wait_with_output()`, which handles this (at the cost of blocking). Instead, it makes repeated calls to to `try_wait()` on timer-based state transitions, and does not try to read the pipes until the worker exits. But when one of the child's pipes is full, the child can block forever waiting on a `write(2)`, such as in a `log` facade implementation.
This bug has not been caught because we control the child worker agent, and until recently, it mostly only wrote to these streams using `env_logger` at its default log level. But recent work: (1) set more-verbose `INFO` level default logging, (2) logged stderr/stdout lines of child processes of _the worker_, and (3) some user targets logged very verbosely for debugging. This surfaced the underlying issue.
