show count of hidden C snapshots

src/signals-mach.c

@@ -550,7 +550,6 @@ void *mach_profile_listener(void *arg)
                 break;
             }
 
-            jl_ptls_t ptls = jl_all_tls_states[i];
             host_thread_state_t state;
             jl_thread_suspend_and_get_state2(i, &state);
             unw_context_t *uc = (unw_context_t*)&state;
@@ -589,6 +588,8 @@ void *mach_profile_listener(void *arg)
 #else
                 bt_size_cur += rec_backtrace_ctx((jl_bt_element_t*)bt_data_prof + bt_size_cur, bt_size_max - bt_size_cur - 1, uc, NULL);
 #endif
+                jl_ptls_t ptls = jl_all_tls_states[i];
+
                 // store threadid but add 1 as 0 is preserved to indicate end of block
                 bt_data_prof[bt_size_cur++].uintptr = ptls->tid + 1;
 

src/signals-unix.c

@@ -749,7 +749,6 @@ static void *signal_listener(void *arg)
         for (int idx = jl_n_threads; idx-- > 0; ) {
             // Stop the threads in the random round-robin order.
             int i = profile_round_robin_thread_order[idx];
-
             // notify thread to stop
             jl_thread_suspend_and_get_state(i, &signal_context);
 

stdlib/Profile/src/Profile.jl

@@ -167,7 +167,7 @@ function print(io::IO,
     else
         if !in(groupby, [:thread, :task, [:task, :thread], [:thread, :task]])
             error(ArgumentError("Unrecognized groupby option: $groupby. Options are :none, :task, :thread, [:task, :thread], or [:thread, :task]"))
-        elseif in(groupby, [:thread, [:task, :thread], [:thread, :task]])
+        elseif Sys.iswindows() && in(groupby, [:thread, [:task, :thread], [:thread, :task]])
             @warn "Profiling on windows is limited to the main thread. Other threads will not show in the report"
         end
         any_nosamples = false
@@ -847,7 +847,8 @@ function tree!(root::StackFrameTree{T}, all::Vector{UInt64}, lidict::Union{LineI
         nothing
     end
     cleanup!(root)
-    return root
+    hidden_count = length(root.down) > 0 ? root.count - mapreduce(x -> x.count, +, values(root.down)) : 0
+    return root, hidden_count
 end
 
 function maxstats(root::StackFrameTree)
@@ -918,24 +919,32 @@ end
 function tree(io::IO, data::Vector{UInt64}, lidict::Union{LineInfoFlatDict, LineInfoDict}, cols::Int, fmt::ProfileFormat,
                 threads::Union{Int,AbstractVector{Int}}, tasks::Union{UInt,AbstractVector{UInt}}, header::Bool, minwarn::Bool)
     if fmt.combine
-        root = tree!(StackFrameTree{StackFrame}(), data, lidict, fmt.C, fmt.recur, threads, tasks)
+        root, hidden_count = tree!(StackFrameTree{StackFrame}(), data, lidict, fmt.C, fmt.recur, threads, tasks)
     else
-        root = tree!(StackFrameTree{UInt64}(), data, lidict, fmt.C, fmt.recur, threads, tasks)
+        root, hidden_count = tree!(StackFrameTree{UInt64}(), data, lidict, fmt.C, fmt.recur, threads, tasks)
     end
     if isempty(root.down)
         if minwarn
             Base.print(io, "Total snapshots: ")
-            printstyled(io, "0\n", color=Base.warn_color())
+            printstyled(io, "$(root.count - hidden_count)", color=Base.warn_color())
+            if hidden_count > 0
+                println(io, " [$(hidden_count) from C hidden]")
+            else
+                println(io)
+            end
         else
             warning_empty()
         end
         return true
     end
     idlecount = print_tree(io, root, cols, fmt, header)
-    Base.print(io, "Total snapshots: ", root.count)
+    Base.print(io, "Total snapshots: ", root.count - hidden_count)
+    if hidden_count > 0
+        Base.print(io, " [$(hidden_count) from C hidden]")
+    end
     if idlecount > 0
         ending = length(threads) > 1 && length(tasks) > 1 ? " across threads $(threads). Use the `groupby` kwarg to separate threads and/or tasks" : ""
-        Base.println(" (", round(Int, (idlecount / root.count) * 100), "% idle$ending)")
+        Base.println(" (", round(Int, (idlecount / (root.count - hidden_count)) * 100), "% idle$ending)")
     else
         Base.println()
     end