support exceptions #50
Labels
enhancement
New feature or request
Comments
|
Would something like below be reasonable ? let[@inline] record_exception (scope : t) (exn : unit -> exn) : unit =
if Collector.has_backend () then (
let ev =
Event.make "exception"
~attrs:
[
"message", `String (Printexc.to_string (exn ()));
"type", `String (Printexc.exn_slot_name (exn ()));
"stacktrace", `String (Printexc.get_backtrace ());
]
in
scope.events <- ev :: scope.events
) |
semgrep-ci bot
pushed a commit
to semgrep/semgrep
that referenced
this issue
Sep 18, 2024
completes SAF-1569 Originally I just wanted to report why semgrep-core was killed in the traces, and send any traces that may have been mid flight when it happened. Since Otel the spec [doesn't support partial traces](open-telemetry/opentelemetry-specification#373), I decided to simply raise the kill signal as an error like we already do when we receive Ctrl c. Otel libraries are supposed to report [exceptions in spans and then exit them](https://opentelemetry.io/docs/specs/semconv/exceptions/exceptions-spans/), but it turns out the otel library [does not do this](imandra-ai/ocaml-opentelemetry#50). So this entire time if there was EVER an error, including file and memory time outs, in a span, we would never send the trace! So this PR makes it so we exit a span whenever an error is caught (and then we reraise it), and the error attributes the library is /supposed/ to add, we add instead! So now we can see what spans have errors, their error message, stacktrace, and we actually see them. ## Test plan ```bash semgrep --pro --trace --config p/default --trace-endpoint semgrep-dev --max-memory 10 --interfile-timeout 1 --timeout 1 --timeout-threshold=1 # wait a bit # then ctrl c or kill -15 <pid> ``` then check jaeger to see traces have error info, and sent traces after ctrl c or kill: [example](https://jaeger-dev2.corp.semgrep.dev/trace/2a86f622e267e7164c29ea0dd37c5286) synced from Pro e8114ab0dccd079b51a0381e4168d5383ab87bbe
semgrep-ci bot
pushed a commit
to semgrep/semgrep
that referenced
this issue
Sep 19, 2024
completes SAF-1569 Originally I just wanted to report why semgrep-core was killed in the traces, and send any traces that may have been mid flight when it happened. Since Otel the spec [doesn't support partial traces](open-telemetry/opentelemetry-specification#373), I decided to simply raise the kill signal as an error like we already do when we receive Ctrl c. Otel libraries are supposed to report [exceptions in spans and then exit them](https://opentelemetry.io/docs/specs/semconv/exceptions/exceptions-spans/), but it turns out the otel library [does not do this](imandra-ai/ocaml-opentelemetry#50). So this entire time if there was EVER an error, including file and memory time outs, in a span, we would never send the trace! So this PR makes it so we exit a span whenever an error is caught (and then we reraise it), and the error attributes the library is /supposed/ to add, we add instead! So now we can see what spans have errors, their error message, stacktrace, and we actually see them. ## Test plan ```bash semgrep --pro --trace --config p/default --trace-endpoint semgrep-dev --max-memory 10 --interfile-timeout 1 --timeout 1 --timeout-threshold=1 # wait a bit # then ctrl c or kill -15 <pid> ``` then check jaeger to see traces have error info, and sent traces after ctrl c or kill: [example](https://jaeger-dev2.corp.semgrep.dev/trace/2a86f622e267e7164c29ea0dd37c5286) synced from Pro e8114ab0dccd079b51a0381e4168d5383ab87bbe
nmote
pushed a commit
to semgrep/semgrep
that referenced
this issue
Sep 19, 2024
completes SAF-1569 Originally I just wanted to report why semgrep-core was killed in the traces, and send any traces that may have been mid flight when it happened. Since Otel the spec [doesn't support partial traces](open-telemetry/opentelemetry-specification#373), I decided to simply raise the kill signal as an error like we already do when we receive Ctrl c. Otel libraries are supposed to report [exceptions in spans and then exit them](https://opentelemetry.io/docs/specs/semconv/exceptions/exceptions-spans/), but it turns out the otel library [does not do this](imandra-ai/ocaml-opentelemetry#50). So this entire time if there was EVER an error, including file and memory time outs, in a span, we would never send the trace! So this PR makes it so we exit a span whenever an error is caught (and then we reraise it), and the error attributes the library is /supposed/ to add, we add instead! So now we can see what spans have errors, their error message, stacktrace, and we actually see them. ## Test plan ```bash semgrep --pro --trace --config p/default --trace-endpoint semgrep-dev --max-memory 10 --interfile-timeout 1 --timeout 1 --timeout-threshold=1 # wait a bit # then ctrl c or kill -15 <pid> ``` then check jaeger to see traces have error info, and sent traces after ctrl c or kill: [example](https://jaeger-dev2.corp.semgrep.dev/trace/2a86f622e267e7164c29ea0dd37c5286) synced from Pro e8114ab0dccd079b51a0381e4168d5383ab87bbe
|
I think this should rather look like: let record_exception (scope:t) (exn:exn) (bt:Printexc.raw_backtrace) : unit =
if …reason is, the exception is never going to be lazy; and afaik |
Merged
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
https://opentelemetry.io/docs/specs/semconv/exceptions/exceptions-spans/
The text was updated successfully, but these errors were encountered: