Merge branch 'main' into 145247-pytuple-from-pair

Author: sergey-miryanov

Doc/Makefile

@@ -88,6 +88,7 @@ htmlhelp: build
 	      "build/htmlhelp/pydoc.hhp project file."
 
 .PHONY: latex
+latex: _ensure-sphinxcontrib-svg2pdfconverter
 latex: BUILDER = latex
 latex: build
 	@echo "Build finished; the LaTeX files are in build/latex."
@@ -231,7 +232,7 @@ dist-text:
 	@echo "Build finished and archived!"
 
 .PHONY: dist-pdf
-dist-pdf:
+dist-pdf: _ensure-sphinxcontrib-svg2pdfconverter
 	# archive the A4 latex
 	@echo "Building LaTeX (A4 paper)..."
 	mkdir -p dist
@@ -292,6 +293,10 @@ _ensure-pre-commit:
 _ensure-sphinx-autobuild:
 	$(MAKE) _ensure-package PACKAGE=sphinx-autobuild
 
+.PHONY: _ensure-sphinxcontrib-svg2pdfconverter
+_ensure-sphinxcontrib-svg2pdfconverter:
+	$(MAKE) _ensure-package PACKAGE=sphinxcontrib-svg2pdfconverter
+
 .PHONY: check
 check: _ensure-pre-commit
 	$(VENVDIR)/bin/python3 -m pre_commit run --all-files

Doc/conf.py

@@ -46,6 +46,7 @@
     'linklint.ext',
     'notfound.extension',
     'sphinxext.opengraph',
+    'sphinxcontrib.rsvgconverter',
 )
 for optional_ext in _OPTIONAL_EXTENSIONS:
     try:

Doc/library/concurrent.futures.rst

@@ -156,7 +156,9 @@ And::
        print(f.result())
 
    executor = ThreadPoolExecutor(max_workers=1)
-   executor.submit(wait_on_future)
+   future = executor.submit(wait_on_future)
+   # Note: calling future.result() would also cause a deadlock because
+   # the single worker thread is already waiting for wait_on_future().
 
 
 .. class:: ThreadPoolExecutor(max_workers=None, thread_name_prefix='', initializer=None, initargs=())

Doc/library/profiling.sampling.rst

@@ -1194,10 +1194,12 @@ data, similar to the ``top`` command for system processes::
    python -m profiling.sampling run --live script.py
    python -m profiling.sampling attach --live 12345
 
-.. figure:: tachyon-live-mode-2.gif
-   :alt: Tachyon live mode showing all threads
-   :align: center
-   :width: 100%
+.. only:: not latex
+
+   .. figure:: tachyon-live-mode-2.gif
+      :alt: Tachyon live mode showing all threads
+      :align: center
+      :width: 100%
 
    Live mode displays real-time profiling statistics, showing combined
    data from multiple threads in a multi-threaded application.
@@ -1217,10 +1219,12 @@ main table, showing instruction-level statistics for the currently selected
 function. This panel displays which bytecode instructions are executing most
 frequently, including specialized variants and their base opcodes.
 
-.. figure:: tachyon-live-mode-1.gif
-   :alt: Tachyon live mode with opcode panel
-   :align: center
-   :width: 100%
+.. only:: not latex
+
+   .. figure:: tachyon-live-mode-1.gif
+      :alt: Tachyon live mode with opcode panel
+      :align: center
+      :width: 100%
 
    Live mode with ``--opcodes`` enabled shows an opcode panel with a bytecode
    instruction breakdown for the selected function.

Python/optimizer.c

@@ -1076,12 +1076,19 @@ _PyJit_FinalizeTracing(PyThreadState *tstate, int err)
             exit->temperature = initial_temperature_backoff_counter(&tstate->interp->opt_config);
         }
     }
+    // Clear all recorded values
+    _PyJitUopBuffer *buffer = &tracer->code_buffer;
+    for (_PyUOpInstruction *inst = buffer->start; inst < buffer->next; inst++) {
+        if (_PyUop_Flags[inst->opcode] & HAS_RECORDS_VALUE_FLAG) {
+            Py_XDECREF((PyObject *)(uintptr_t)inst->operand0);
+        }
+    }
     Py_CLEAR(tracer->initial_state.code);
     Py_CLEAR(tracer->initial_state.func);
     Py_CLEAR(tracer->initial_state.executor);
     Py_CLEAR(tracer->prev_state.instr_code);
     Py_CLEAR(tracer->prev_state.recorded_value);
-    uop_buffer_init(&tracer->code_buffer, &tracer->uop_array[0], UOP_MAX_TRACE_LENGTH);
+    uop_buffer_init(buffer, &tracer->uop_array[0], UOP_MAX_TRACE_LENGTH);
     tracer->is_tracing = false;
 }
 
@@ -1521,6 +1528,11 @@ uop_optimize(
         }
         assert(_PyOpcode_uop_name[buffer[pc].opcode]);
     }
+    // We've cleaned up the references in the buffer, so discard the code buffer
+    // to avoid doing it again during tracer cleanup
+    _PyJitUopBuffer *code_buffer = &_tstate->jit_tracer_state->code_buffer;
+    code_buffer->next = code_buffer->start;
+
     OPT_HIST(effective_trace_length(buffer, length), optimized_trace_length_hist);
     _PyUOpInstruction *output = &_tstate->jit_tracer_state->uop_array[0];
     length = stack_allocate(buffer, output, length);