0.8

add trace.subscriber instead of a separate library
add trace-tef.tldrs, to trace multiple processes easily (with external rust daemon)
breaking: trace-tef: use mtime.now, not a counter, for multiproc
trace-fuchsia: require thread-local-storage 0.2

0.7

feat: add levels to Trace_core. Levels are similar to logs levels, to help control verbosity.
add hmap as a depopt (#28)
fix: truncate large strings in fuchsia

0.6

add ppx_trace for easier instrumentation.
- let%trace span = "foo" in … will enter a scope span named "foo"
- let%trace () = "foo" in … will enter a scope named "foo" with a hidden name
add trace-fuchsia backend, which produces traces in the binary format of fuchsia. These traces are reasonably efficient to produce (~60ns per span on my machines) and reasonably compact on disk, at least compared to the TEF backend.

0.5

perf: reduce overhead in trace-tef
perf: add Mpsc_queue, adapted from picos, to trace-tef

0.4

add ?data to counter_int and counter_float
add float to user data
add add_data_to_current_span and add_data_to_manual_span
make explicit_span.meta mutable
trace-tef: write to trace.json if env variable TRACE is either 1 or true
trace-tef: emit function name, if provided, as a metadata key/value pair
re-export trace.core in trace
perf: in trace-tef, use broadcast instead of signal in the job queue

0.3

add explicit spans, for more precise tracing
rename repo to ocaml-trace
trace-tef: add a ticker thread to ensure we flush the file regularly

0.2

trace-tef: additional argument to with_setup; env for "stdout"/"stderr"
refactor: avoid conflicting with stdlib Trace module by adding sublibrary trace.core. Programs that use compiler-libs.toplevel should use trace.core directly, because using trace will cause linking errors.
perf(trace-tef): improve behavior of collector under contention by pulling all events at once in the worker

0.1

initial release