With Flowtrace we have a way to persist traces of a network execution. fbp-spec should make use of this to allow debugging after failing test runs.
For each case, runner should set up tracing using trace:start. Then if test fails, trigger a trace:dump to retrieve it.
Should possibly also have an option to store trace even if test passes (per testcase). Trace from successful run can be useful to compare against a failure, or just to understand the program in general.
The CLI should maybe just dump these files to disk, with some meaningful names (derived from testcase names?). Can then use regular Flowtrace tools with them. Alternatively allow them to be uploaded somewhere (useful for CI).
For the browser/interactive case, where we're running in Flowhub or similar, must be possible to access the traces - so that the data can be loaded into editor/debugger for inspection.
With Flowtrace we have a way to persist traces of a network execution. fbp-spec should make use of this to allow debugging after failing test runs.
For each case, runner should set up tracing using
trace:start
. Then if test fails, trigger atrace:dump
to retrieve it. Should possibly also have an option to store trace even if test passes (per testcase). Trace from successful run can be useful to compare against a failure, or just to understand the program in general.The CLI should maybe just dump these files to disk, with some meaningful names (derived from testcase names?). Can then use regular Flowtrace tools with them. Alternatively allow them to be uploaded somewhere (useful for CI).