allow getting task dump backtraces in a programmatic format #6975
Conversation
github-actions
bot
added
R-loom-current-thread
arielb1
force-pushed
the
programmatic-backtraces
branch
3 times, most recently
from
df04fc7
to
ede6e06
Compare
|
What is this waiting for? |
Darksonn
added
A-tokio
tokio/src/runtime/dump.rs
Outdated
| /// Return the raw name of the symbol. | ||
| pub fn name_raw(&self) -> Option<&[u8]> { | ||
| self.name.as_deref() | ||
| } | ||
|
|
||
| /// Return the demangled name of the symbol. | ||
| pub fn name_demangled(&self) -> Option<&str> { | ||
| self.name_demangled.as_deref() | ||
| } |
There was a problem hiding this comment.
Do we need both? I think users can construct demangled one by using name_raw.
There was a problem hiding this comment.
I just had it since backtrace had that, and for some reason I thought the API was somewhat private. Apparently rustc_demangle is a perfectly fine public API.
tokio/src/runtime/dump.rs
Outdated
| /// Resolve and return a list of backtraces that are involved in polls in this task. | ||
| pub fn resolve_backtraces(&self) -> Vec<Vec<BacktraceFrame>> { |
There was a problem hiding this comment.
Is it possible for one Trace to return multiple backtraces? If such a case does not exist, I feel it would be clearer to return Vec<BacktraceFrame>.
There was a problem hiding this comment.
I suspect this may have to do with tokio::join! and similar.
There was a problem hiding this comment.
Added documentation to that effect.
| let mut backtrace = backtrace::Backtrace::from(backtrace.clone()); | ||
| backtrace.resolve(); | ||
| backtrace | ||
| .frames() | ||
| .iter() | ||
| .map(BacktraceFrame::from_resolved_backtrace_frame) | ||
| .collect() |
There was a problem hiding this comment.
Perhaps we can fix this in the follow up PR, but if we are going to have our own Backtrace type, I think we can avoid this conversion by making trace_leaf function construct tokio's Backtrace struct directly.
There was a problem hiding this comment.
this resolves the backtrace, and therefore is expensive (see warning in docs), and therefore shouldn't be done in trace_leaf
arielb1
force-pushed
the
programmatic-backtraces
branch
from
0521d0a
to
283eea6
Compare
arielb1
force-pushed
the
programmatic-backtraces
branch
from
283eea6
to
dd3f267
Compare
This PR creates structs that allow dumping backtraces in a programmatic format, for example JSON. A JSON serializer is not natively included to avoid creating a Tokio dependency on one, but serialization can be done by customers.
Fixes #6309.