[analyze-memory] Add RAM symbol analysis by component (#13040)

esphome/__main__.py

@@ -1017,6 +1017,7 @@ def command_analyze_memory(args: ArgsProtocol, config: ConfigType) -> int:
         idedata.objdump_path,
         idedata.readelf_path,
         external_components,
+        idedata=idedata,
     )
     analyzer.analyze()
 

esphome/analyze_memory/__init__.py

@@ -22,6 +22,7 @@ from .helpers import (
     map_section_name,
     parse_symbol_line,
 )
+from .toolchain import find_tool, run_tool
 
 if TYPE_CHECKING:
     from esphome.platformio_api import IDEData
@@ -53,6 +54,9 @@ _NAMESPACE_STD = "std::"
 # Type alias for symbol information: (symbol_name, size, component)
 SymbolInfoType = tuple[str, int, str]
 
+# RAM sections - symbols in these sections consume RAM
+RAM_SECTIONS = frozenset([".data", ".bss"])
+
 
 @dataclass
 class MemorySection:
@@ -60,7 +64,20 @@ class MemorySection:
 
     name: str
     symbols: list[SymbolInfoType] = field(default_factory=list)
-    total_size: int = 0
+    total_size: int = 0  # Actual section size from ELF headers
+    symbol_size: int = 0  # Sum of symbol sizes (may be less than total_size)
+
+
+@dataclass
+class SDKSymbol:
+    """Represents a symbol from an SDK library that's not in the ELF symbol table."""
+
+    name: str
+    size: int
+    library: str  # Name of the .a file (e.g., "libpp.a")
+    section: str  # ".bss" or ".data"
+    is_local: bool  # True if static/local symbol (lowercase in nm output)
+    demangled: str = ""  # Demangled name (populated after analysis)
 
 
 @dataclass
@@ -118,6 +135,10 @@ class MemoryAnalyzer:
         self.objdump_path = objdump_path or "objdump"
         self.readelf_path = readelf_path or "readelf"
         self.external_components = external_components or set()
+        self._idedata = idedata
+
+        # Derive nm path from objdump path using shared toolchain utility
+        self.nm_path = find_tool("nm", self.objdump_path)
 
         self.sections: dict[str, MemorySection] = {}
         self.components: dict[str, ComponentMemory] = defaultdict(
@@ -128,15 +149,25 @@ class MemoryAnalyzer:
         self._esphome_core_symbols: list[
             tuple[str, str, int]
         ] = []  # Track core symbols
-        self._component_symbols: dict[str, list[tuple[str, str, int]]] = defaultdict(
+        # Track symbols for all components: (symbol_name, demangled, size, section)
+        self._component_symbols: dict[str, list[tuple[str, str, int, str]]] = (
+            defaultdict(list)
+        )
+        # Track RAM symbols separately for detailed analysis: (symbol_name, demangled, size, section)
+        self._ram_symbols: dict[str, list[tuple[str, str, int, str]]] = defaultdict(
             list
-        )  # Track symbols for all components
+        )
+        # Track ELF symbol names for SDK cross-reference
+        self._elf_symbol_names: set[str] = set()
+        # SDK symbols not in ELF (static/local symbols from closed-source libs)
+        self._sdk_symbols: list[SDKSymbol] = []
 
     def analyze(self) -> dict[str, ComponentMemory]:
         """Analyze the ELF file and return component memory usage."""
         self._parse_sections()
         self._parse_symbols()
         self._categorize_symbols()
+        self._analyze_sdk_libraries()
         return dict(self.components)
 
     def _parse_sections(self) -> None:
@@ -190,6 +221,8 @@ class MemoryAnalyzer:
                 continue
 
             self.sections[section].symbols.append((name, size, ""))
+            self.sections[section].symbol_size += size
+            self._elf_symbol_names.add(name)
             seen_addresses.add(address)
 
     def _categorize_symbols(self) -> None:
@@ -233,8 +266,13 @@ class MemoryAnalyzer:
                 if size > 0:
                     demangled = self._demangle_symbol(symbol_name)
                     self._component_symbols[component].append(
-                        (symbol_name, demangled, size)
+                        (symbol_name, demangled, size, section_name)
                     )
+                    # Track RAM symbols separately for detailed RAM analysis
+                    if section_name in RAM_SECTIONS:
+                        self._ram_symbols[component].append(
+                            (symbol_name, demangled, size, section_name)
+                        )
 
     def _identify_component(self, symbol_name: str) -> str:
         """Identify which component a symbol belongs to."""
@@ -328,6 +366,247 @@ class MemoryAnalyzer:
 
         return "Other Core"
 
+    def get_unattributed_ram(self) -> tuple[int, int, int]:
+        """Get unattributed RAM sizes (SDK/framework overhead).
+
+        Returns:
+            Tuple of (unattributed_bss, unattributed_data, total_unattributed)
+            These are bytes in RAM sections that have no corresponding symbols.
+        """
+        bss_section = self.sections.get(".bss")
+        data_section = self.sections.get(".data")
+
+        unattributed_bss = 0
+        unattributed_data = 0
+
+        if bss_section:
+            unattributed_bss = max(0, bss_section.total_size - bss_section.symbol_size)
+        if data_section:
+            unattributed_data = max(
+                0, data_section.total_size - data_section.symbol_size
+            )
+
+        return unattributed_bss, unattributed_data, unattributed_bss + unattributed_data
+
+    def _find_sdk_library_dirs(self) -> list[Path]:
+        """Find SDK library directories based on platform.
+
+        Returns:
+            List of paths to SDK library directories containing .a files.
+        """
+        sdk_dirs: list[Path] = []
+
+        if self._idedata is None:
+            return sdk_dirs
+
+        # Get the CC path to determine the framework location
+        cc_path = getattr(self._idedata, "cc_path", None)
+        if not cc_path:
+            return sdk_dirs
+
+        cc_path = Path(cc_path)
+
+        # For ESP8266 Arduino framework
+        # CC is like: ~/.platformio/packages/toolchain-xtensa/bin/xtensa-lx106-elf-gcc
+        # SDK libs are in: ~/.platformio/packages/framework-arduinoespressif8266/tools/sdk/lib/
+        if "xtensa-lx106" in str(cc_path):
+            platformio_dir = cc_path.parent.parent.parent
+            esp8266_sdk = (
+                platformio_dir
+                / "framework-arduinoespressif8266"
+                / "tools"
+                / "sdk"
+                / "lib"
+            )
+            if esp8266_sdk.exists():
+                sdk_dirs.append(esp8266_sdk)
+                # Also check for NONOSDK subdirectories (closed-source libs)
+                sdk_dirs.extend(
+                    subdir
+                    for subdir in esp8266_sdk.iterdir()
+                    if subdir.is_dir() and subdir.name.startswith("NONOSDK")
+                )
+
+        # For ESP32 IDF framework
+        # CC is like: ~/.platformio/packages/toolchain-xtensa-esp-elf/bin/xtensa-esp32-elf-gcc
+        # or: ~/.platformio/packages/toolchain-riscv32-esp/bin/riscv32-esp-elf-gcc
+        elif "xtensa-esp" in str(cc_path) or "riscv32-esp" in str(cc_path):
+            # Detect ESP32 variant from CC path or defines
+            variant = self._detect_esp32_variant()
+            if variant:
+                platformio_dir = cc_path.parent.parent.parent
+                espidf_dir = platformio_dir / "framework-espidf" / "components"
+                if espidf_dir.exists():
+                    # Find all directories named after the variant that contain .a files
+                    # This handles various ESP-IDF library layouts:
+                    # - components/*/lib/<variant>/
+                    # - components/*/<variant>/
+                    # - components/*/lib/lib/<variant>/
+                    # - components/*/*/lib_*/<variant>/
+                    sdk_dirs.extend(
+                        variant_dir
+                        for variant_dir in espidf_dir.rglob(variant)
+                        if variant_dir.is_dir() and any(variant_dir.glob("*.a"))
+                    )
+
+        return sdk_dirs
+
+    def _detect_esp32_variant(self) -> str | None:
+        """Detect ESP32 variant from idedata defines.
+
+        Returns:
+            Variant string like 'esp32', 'esp32s2', 'esp32c3', etc. or None.
+        """
+        if self._idedata is None:
+            return None
+
+        defines = getattr(self._idedata, "defines", [])
+        if not defines:
+            return None
+
+        # ESPHome always adds USE_ESP32_VARIANT_xxx defines
+        variant_prefix = "USE_ESP32_VARIANT_"
+        for define in defines:
+            if define.startswith(variant_prefix):
+                # Extract variant name and convert to lowercase
+                # USE_ESP32_VARIANT_ESP32 -> esp32
+                # USE_ESP32_VARIANT_ESP32S3 -> esp32s3
+                return define[len(variant_prefix) :].lower()
+
+        return None
+
+    def _parse_sdk_library(
+        self, lib_path: Path
+    ) -> tuple[list[tuple[str, int, str, bool]], set[str]]:
+        """Parse a single SDK library for symbols.
+
+        Args:
+            lib_path: Path to the .a library file
+
+        Returns:
+            Tuple of:
+            - List of BSS/DATA symbols: (symbol_name, size, section, is_local)
+            - Set of global BSS/DATA symbol names (for checking if RAM is linked)
+        """
+        ram_symbols: list[tuple[str, int, str, bool]] = []
+        global_ram_symbols: set[str] = set()
+
+        result = run_tool([self.nm_path, "--size-sort", str(lib_path)], timeout=10)
+        if result is None:
+            return ram_symbols, global_ram_symbols
+
+        for line in result.stdout.splitlines():
+            parts = line.split()
+            if len(parts) < 3:
+                continue
+
+            try:
+                size = int(parts[0], 16)
+                sym_type = parts[1]
+                name = parts[2]
+
+                # Only collect BSS (b/B) and DATA (d/D) for RAM analysis
+                if sym_type in ("b", "B"):
+                    section = ".bss"
+                    is_local = sym_type == "b"
+                    ram_symbols.append((name, size, section, is_local))
+                    # Track global RAM symbols (B/D) for linking check
+                    if sym_type == "B":
+                        global_ram_symbols.add(name)
+                elif sym_type in ("d", "D"):
+                    section = ".data"
+                    is_local = sym_type == "d"
+                    ram_symbols.append((name, size, section, is_local))
+                    if sym_type == "D":
+                        global_ram_symbols.add(name)
+            except (ValueError, IndexError):
+                continue
+
+        return ram_symbols, global_ram_symbols
+
+    def _analyze_sdk_libraries(self) -> None:
+        """Analyze SDK libraries to find symbols not in the ELF.
+
+        This finds static/local symbols from closed-source SDK libraries
+        that consume RAM but don't appear in the final ELF symbol table.
+        Only includes symbols from libraries that have RAM actually linked
+        (at least one global BSS/DATA symbol in the ELF).
+        """
+        sdk_dirs = self._find_sdk_library_dirs()
+        if not sdk_dirs:
+            _LOGGER.debug("No SDK library directories found")
+            return
+
+        _LOGGER.debug("Analyzing SDK libraries in %d directories", len(sdk_dirs))
+
+        # Track seen symbols to avoid duplicates from multiple SDK versions
+        seen_symbols: set[str] = set()
+
+        for sdk_dir in sdk_dirs:
+            for lib_path in sorted(sdk_dir.glob("*.a")):
+                lib_name = lib_path.name
+                ram_symbols, global_ram_symbols = self._parse_sdk_library(lib_path)
+
+                # Check if this library's RAM is actually linked by seeing if any
+                # of its global BSS/DATA symbols appear in the ELF
+                if not global_ram_symbols & self._elf_symbol_names:
+                    # No RAM from this library is in the ELF - skip it
+                    continue
+
+                for name, size, section, is_local in ram_symbols:
+                    # Skip if already in ELF or already seen from another lib
+                    if name in self._elf_symbol_names or name in seen_symbols:
+                        continue
+
+                    # Only track symbols with non-zero size
+                    if size > 0:
+                        self._sdk_symbols.append(
+                            SDKSymbol(
+                                name=name,
+                                size=size,
+                                library=lib_name,
+                                section=section,
+                                is_local=is_local,
+                            )
+                        )
+                        seen_symbols.add(name)
+
+        # Demangle SDK symbols for better readability
+        if self._sdk_symbols:
+            sdk_names = [sym.name for sym in self._sdk_symbols]
+            demangled_map = batch_demangle(sdk_names, objdump_path=self.objdump_path)
+            for sym in self._sdk_symbols:
+                sym.demangled = demangled_map.get(sym.name, sym.name)
+
+        # Sort by size descending for reporting
+        self._sdk_symbols.sort(key=lambda s: s.size, reverse=True)
+
+        total_sdk_ram = sum(s.size for s in self._sdk_symbols)
+        _LOGGER.debug(
+            "Found %d SDK symbols not in ELF, totaling %d bytes",
+            len(self._sdk_symbols),
+            total_sdk_ram,
+        )
+
+    def get_sdk_ram_symbols(self) -> list[SDKSymbol]:
+        """Get SDK symbols that consume RAM but aren't in the ELF symbol table.
+
+        Returns:
+            List of SDKSymbol objects sorted by size descending.
+        """
+        return self._sdk_symbols
+
+    def get_sdk_ram_by_library(self) -> dict[str, list[SDKSymbol]]:
+        """Get SDK RAM symbols grouped by library.
+
+        Returns:
+            Dictionary mapping library name to list of symbols.
+        """
+        by_lib: dict[str, list[SDKSymbol]] = defaultdict(list)
+        for sym in self._sdk_symbols:
+            by_lib[sym.library].append(sym)
+        return dict(by_lib)
+
 
 if __name__ == "__main__":
     from .cli import main

esphome/analyze_memory/cli.py

@@ -1,16 +1,24 @@
 """CLI interface for memory analysis with report generation."""
 
+from __future__ import annotations
+
 from collections import defaultdict
+from collections.abc import Callable
 import sys
+from typing import TYPE_CHECKING
 
 from . import (
     _COMPONENT_API,
     _COMPONENT_CORE,
     _COMPONENT_PREFIX_ESPHOME,
     _COMPONENT_PREFIX_EXTERNAL,
+    RAM_SECTIONS,
     MemoryAnalyzer,
 )
 
+if TYPE_CHECKING:
+    from . import ComponentMemory
+
 
 class MemoryAnalyzerCLI(MemoryAnalyzer):
     """Memory analyzer with CLI-specific report generation."""
@@ -19,6 +27,8 @@ class MemoryAnalyzerCLI(MemoryAnalyzer):
     SYMBOL_SIZE_THRESHOLD: int = (
         100  # Show symbols larger than this in detailed analysis
     )
+    # Lower threshold for RAM symbols (RAM is more constrained)
+    RAM_SYMBOL_SIZE_THRESHOLD: int = 24
 
     # Column width constants
     COL_COMPONENT: int = 29
@@ -83,6 +93,60 @@ class MemoryAnalyzerCLI(MemoryAnalyzer):
         COL_CORE_PERCENT,
     )
 
+    def _add_section_header(self, lines: list[str], title: str) -> None:
+        """Add a section header with title centered between separator lines."""
+        lines.append("")
+        lines.append("=" * self.TABLE_WIDTH)
+        lines.append(title.center(self.TABLE_WIDTH))
+        lines.append("=" * self.TABLE_WIDTH)
+        lines.append("")
+
+    def _add_top_consumers(
+        self,
+        lines: list[str],
+        title: str,
+        components: list[tuple[str, ComponentMemory]],
+        get_size: Callable[[ComponentMemory], int],
+        total: int,
+        memory_type: str,
+        limit: int = 25,
+    ) -> None:
+        """Add a formatted list of top memory consumers to the report.
+
+        Args:
+            lines: List of report lines to append the output to.
+            title: Section title to print before the list.
+            components: Sequence of (name, ComponentMemory) tuples to analyze.
+            get_size: Callable that takes a ComponentMemory and returns the
+                size in bytes to use for ranking and display.
+            total: Total size in bytes for computing percentage usage.
+            memory_type: Label for the memory region (e.g., "flash" or "RAM").
+            limit: Maximum number of components to include in the list.
+        """
+        lines.append("")
+        lines.append(f"{title}:")
+        for i, (name, mem) in enumerate(components[:limit]):
+            size = get_size(mem)
+            if size > 0:
+                percentage = (size / total * 100) if total > 0 else 0
+                lines.append(
+                    f"{i + 1}. {name} ({size:,} B) - {percentage:.1f}% of analyzed {memory_type}"
+                )
+
+    def _format_symbol_with_section(
+        self, demangled: str, size: int, section: str | None = None
+    ) -> str:
+        """Format a symbol entry, optionally adding a RAM section label.
+
+        If section is one of the RAM sections (.data or .bss), a label like
+        " [data]" or " [bss]" is appended. For non-RAM sections or when
+        section is None, no section label is added.
+        """
+        section_label = ""
+        if section in RAM_SECTIONS:
+            section_label = f" [{section[1:]}]"  # .data -> [data], .bss -> [bss]
+        return f"{demangled} ({size:,} B){section_label}"
+
     def generate_report(self, detailed: bool = False) -> str:
         """Generate a formatted memory report."""
         components = sorted(
@@ -123,43 +187,70 @@ class MemoryAnalyzerCLI(MemoryAnalyzer):
             f"{total_flash:>{self.COL_TOTAL_FLASH - 2},} B | {total_ram:>{self.COL_TOTAL_RAM - 2},} B"
         )
 
-        # Top consumers
-        lines.append("")
-        lines.append("Top Flash Consumers:")
-        for i, (name, mem) in enumerate(components[:25]):
-            if mem.flash_total > 0:
-                percentage = (
-                    (mem.flash_total / total_flash * 100) if total_flash > 0 else 0
-                )
-                lines.append(
-                    f"{i + 1}. {name} ({mem.flash_total:,} B) - {percentage:.1f}% of analyzed flash"
-                )
-
-        lines.append("")
-        lines.append("Top RAM Consumers:")
-        ram_components = sorted(components, key=lambda x: x[1].ram_total, reverse=True)
-        for i, (name, mem) in enumerate(ram_components[:25]):
-            if mem.ram_total > 0:
-                percentage = (mem.ram_total / total_ram * 100) if total_ram > 0 else 0
-                lines.append(
-                    f"{i + 1}. {name} ({mem.ram_total:,} B) - {percentage:.1f}% of analyzed RAM"
-                )
-
-        lines.append("")
-        lines.append(
-            "Note: This analysis covers symbols in the ELF file. Some runtime allocations may not be included."
+        # Show unattributed RAM (SDK/framework overhead)
+        unattributed_bss, unattributed_data, unattributed_total = (
+            self.get_unattributed_ram()
+        )
+        if unattributed_total > 0:
+            lines.append("")
+            lines.append(
+                f"Unattributed RAM: {unattributed_total:,} B (SDK/framework overhead)"
+            )
+            if unattributed_bss > 0 and unattributed_data > 0:
+                lines.append(
+                    f"  .bss: {unattributed_bss:,} B | .data: {unattributed_data:,} B"
+                )
+
+            # Show SDK symbol breakdown if available
+            sdk_by_lib = self.get_sdk_ram_by_library()
+            if sdk_by_lib:
+                lines.append("")
+                lines.append("SDK library breakdown (static symbols not in ELF):")
+                # Sort libraries by total size
+                lib_totals = [
+                    (lib, sum(s.size for s in syms), syms)
+                    for lib, syms in sdk_by_lib.items()
+                ]
+                lib_totals.sort(key=lambda x: x[1], reverse=True)
+
+                for lib_name, lib_total, syms in lib_totals:
+                    if lib_total == 0:
+                        continue
+                    lines.append(f"  {lib_name}: {lib_total:,} B")
+                    # Show top symbols from this library
+                    for sym in sorted(syms, key=lambda s: s.size, reverse=True)[:3]:
+                        section_label = sym.section.lstrip(".")
+                        # Use demangled name (falls back to original if not demangled)
+                        display_name = sym.demangled or sym.name
+                        if len(display_name) > 50:
+                            display_name = f"{display_name[:47]}..."
+                        lines.append(
+                            f"    {sym.size:>6,} B [{section_label}] {display_name}"
+                        )
+
+        # Top consumers
+        self._add_top_consumers(
+            lines,
+            "Top Flash Consumers",
+            components,
+            lambda m: m.flash_total,
+            total_flash,
+            "flash",
+        )
+
+        ram_components = sorted(components, key=lambda x: x[1].ram_total, reverse=True)
+        self._add_top_consumers(
+            lines,
+            "Top RAM Consumers",
+            ram_components,
+            lambda m: m.ram_total,
+            total_ram,
+            "RAM",
         )
-        lines.append("=" * self.TABLE_WIDTH)
 
         # Add ESPHome core detailed analysis if there are core symbols
         if self._esphome_core_symbols:
-            lines.append("")
-            lines.append("=" * self.TABLE_WIDTH)
-            lines.append(
-                f"{_COMPONENT_CORE} Detailed Analysis".center(self.TABLE_WIDTH)
-            )
-            lines.append("=" * self.TABLE_WIDTH)
-            lines.append("")
+            self._add_section_header(lines, f"{_COMPONENT_CORE} Detailed Analysis")
 
             # Group core symbols by subcategory
             core_subcategories: dict[str, list[tuple[str, str, int]]] = defaultdict(
@@ -211,7 +302,11 @@ class MemoryAnalyzerCLI(MemoryAnalyzer):
                 f"{_COMPONENT_CORE} Symbols > {self.SYMBOL_SIZE_THRESHOLD} B ({len(large_core_symbols)} symbols):"
             )
             for i, (symbol, demangled, size) in enumerate(large_core_symbols):
-                lines.append(f"{i + 1}. {demangled} ({size:,} B)")
+                # Core symbols only track (symbol, demangled, size) without section info,
+                # so we don't show section labels here
+                lines.append(
+                    f"{i + 1}. {self._format_symbol_with_section(demangled, size)}"
+                )
 
             lines.append("=" * self.TABLE_WIDTH)
 
@@ -267,11 +362,7 @@ class MemoryAnalyzerCLI(MemoryAnalyzer):
             for comp_name, comp_mem in components_to_analyze:
                 if not (comp_symbols := self._component_symbols.get(comp_name, [])):
                     continue
-                lines.append("")
-                lines.append("=" * self.TABLE_WIDTH)
-                lines.append(f"{comp_name} Detailed Analysis".center(self.TABLE_WIDTH))
-                lines.append("=" * self.TABLE_WIDTH)
-                lines.append("")
+                self._add_section_header(lines, f"{comp_name} Detailed Analysis")
 
                 # Sort symbols by size
                 sorted_symbols = sorted(comp_symbols, key=lambda x: x[2], reverse=True)
@@ -282,19 +373,69 @@ class MemoryAnalyzerCLI(MemoryAnalyzer):
 
                 # Show all symbols above threshold for better visibility
                 large_symbols = [
-                    (sym, dem, size)
-                    for sym, dem, size in sorted_symbols
+                    (sym, dem, size, sec)
+                    for sym, dem, size, sec in sorted_symbols
                     if size > self.SYMBOL_SIZE_THRESHOLD
                 ]
 
                 lines.append(
                     f"{comp_name} Symbols > {self.SYMBOL_SIZE_THRESHOLD} B ({len(large_symbols)} symbols):"
                 )
-                for i, (symbol, demangled, size) in enumerate(large_symbols):
-                    lines.append(f"{i + 1}. {demangled} ({size:,} B)")
+                for i, (symbol, demangled, size, section) in enumerate(large_symbols):
+                    lines.append(
+                        f"{i + 1}. {self._format_symbol_with_section(demangled, size, section)}"
+                    )
 
                 lines.append("=" * self.TABLE_WIDTH)
 
+        # Detailed RAM analysis by component (at end, before RAM strings analysis)
+        self._add_section_header(lines, "RAM Symbol Analysis by Component")
+
+        # Show top 15 RAM consumers with their large symbols
+        for name, mem in ram_components[:15]:
+            if mem.ram_total == 0:
+                continue
+            ram_syms = self._ram_symbols.get(name, [])
+            if not ram_syms:
+                continue
+
+            # Sort by size descending
+            sorted_ram_syms = sorted(ram_syms, key=lambda x: x[2], reverse=True)
+            large_ram_syms = [
+                s for s in sorted_ram_syms if s[2] > self.RAM_SYMBOL_SIZE_THRESHOLD
+            ]
+
+            lines.append(f"{name} ({mem.ram_total:,} B total RAM):")
+
+            # Show breakdown by section type
+            data_size = sum(s[2] for s in ram_syms if s[3] == ".data")
+            bss_size = sum(s[2] for s in ram_syms if s[3] == ".bss")
+            lines.append(f"  .data (initialized): {data_size:,} B")
+            lines.append(f"  .bss (uninitialized): {bss_size:,} B")
+
+            if large_ram_syms:
+                lines.append(
+                    f"  Symbols > {self.RAM_SYMBOL_SIZE_THRESHOLD} B ({len(large_ram_syms)}):"
+                )
+                for symbol, demangled, size, section in large_ram_syms[:10]:
+                    # Format section label consistently by stripping leading dot
+                    section_label = section.lstrip(".") if section else ""
+                    # Add ellipsis if name is truncated
+                    demangled_display = (
+                        f"{demangled[:70]}..." if len(demangled) > 70 else demangled
+                    )
+                    lines.append(
+                        f"    {size:>6,} B [{section_label}] {demangled_display}"
+                    )
+                if len(large_ram_syms) > 10:
+                    lines.append(f"    ... and {len(large_ram_syms) - 10} more")
+            lines.append("")
+
+        lines.append(
+            "Note: This analysis covers symbols in the ELF file. Some runtime allocations may not be included."
+        )
+        lines.append("=" * self.TABLE_WIDTH)
+
         return "\n".join(lines)
 
     def dump_uncategorized_symbols(self, output_file: str | None = None) -> None:

esphome/analyze_memory/const.py

@@ -7,11 +7,13 @@ ESPHOME_COMPONENT_PATTERN = re.compile(r"esphome::([a-zA-Z0-9_]+)::")
 
 # Section mapping for ELF file sections
 # Maps standard section names to their various platform-specific variants
+# Note: Order matters! More specific patterns (.bss) must come before general ones (.dram)
+# because ESP-IDF uses names like ".dram0.bss" which would match ".dram" otherwise
 SECTION_MAPPING = {
     ".text": frozenset([".text", ".iram"]),
     ".rodata": frozenset([".rodata"]),
+    ".bss": frozenset([".bss"]),  # Must be before .data to catch ".dram0.bss"
     ".data": frozenset([".data", ".dram"]),
-    ".bss": frozenset([".bss"]),
 }
 
 # Section to ComponentMemory attribute mapping

esphome/analyze_memory/toolchain.py

@@ -5,6 +5,10 @@ from __future__ import annotations
 import logging
 from pathlib import Path
 import subprocess
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from collections.abc import Sequence
 
 _LOGGER = logging.getLogger(__name__)
 
@@ -55,3 +59,35 @@ def find_tool(
 
     _LOGGER.warning("Could not find %s tool", tool_name)
     return None
+
+
+def run_tool(
+    cmd: Sequence[str],
+    timeout: int = 30,
+) -> subprocess.CompletedProcess[str] | None:
+    """Run a toolchain command and return the result.
+
+    Args:
+        cmd: Command and arguments to run
+        timeout: Timeout in seconds
+
+    Returns:
+        CompletedProcess on success, None on failure
+    """
+    try:
+        return subprocess.run(
+            cmd,
+            capture_output=True,
+            text=True,
+            timeout=timeout,
+            check=False,
+        )
+    except subprocess.TimeoutExpired:
+        _LOGGER.warning("Command timed out: %s", " ".join(cmd))
+        return None
+    except FileNotFoundError:
+        _LOGGER.warning("Command not found: %s", cmd[0])
+        return None
+    except OSError as e:
+        _LOGGER.warning("Failed to run command %s: %s", cmd[0], e)
+        return None

esphome/platformio_api.py

@@ -420,3 +420,8 @@ class IDEData:
             if path.endswith(".exe")
             else f"{path[:-3]}readelf"
         )
+
+    @property
+    def defines(self) -> list[str]:
+        """Return the list of preprocessor defines from idedata."""
+        return self.raw.get("defines", [])

tests/unit_tests/test_main.py

@@ -2478,6 +2478,7 @@ def test_command_analyze_memory_success(
         "/path/to/objdump",
         "/path/to/readelf",
         set(),  # No external components
+        idedata=mock_get_idedata.return_value,
     )
 
     # Verify analysis was run
@@ -2547,6 +2548,7 @@ def test_command_analyze_memory_with_external_components(
         "/path/to/objdump",
         "/path/to/readelf",
         {"my_custom_component"},  # External component detected
+        idedata=mock_get_idedata.return_value,
     )