Files
esphome/esphome/analyze_memory/helpers.py
J. Nick Koston 4f4da1de22 preen
2025-10-17 18:41:12 -10:00

122 lines
3.6 KiB
Python

"""Helper functions for memory analysis."""
from functools import cache
from pathlib import Path
from .const import SECTION_MAPPING
# Import namespace constant from parent module
# Note: This would create a circular import if done at module level,
# so we'll define it locally here as well
_NAMESPACE_ESPHOME = "esphome::"
# Get the list of actual ESPHome components by scanning the components directory
@cache
def get_esphome_components():
"""Get set of actual ESPHome components from the components directory."""
# Find the components directory relative to this file
# Go up two levels from analyze_memory/helpers.py to esphome/
current_dir = Path(__file__).parent.parent
components_dir = current_dir / "components"
if not components_dir.exists() or not components_dir.is_dir():
return frozenset()
return frozenset(
item.name
for item in components_dir.iterdir()
if item.is_dir()
and not item.name.startswith(".")
and not item.name.startswith("__")
)
@cache
def get_component_class_patterns(component_name: str) -> list[str]:
"""Generate component class name patterns for symbol matching.
Args:
component_name: The component name (e.g., "ota", "wifi", "api")
Returns:
List of pattern strings to match against demangled symbols
"""
component_upper = component_name.upper()
component_camel = component_name.replace("_", "").title()
return [
f"{_NAMESPACE_ESPHOME}{component_upper}Component", # e.g., esphome::OTAComponent
f"{_NAMESPACE_ESPHOME}ESPHome{component_upper}Component", # e.g., esphome::ESPHomeOTAComponent
f"{_NAMESPACE_ESPHOME}{component_camel}Component", # e.g., esphome::OtaComponent
f"{_NAMESPACE_ESPHOME}ESPHome{component_camel}Component", # e.g., esphome::ESPHomeOtaComponent
]
def map_section_name(raw_section: str) -> str | None:
"""Map raw section name to standard section.
Args:
raw_section: Raw section name from ELF file (e.g., ".iram0.text", ".rodata.str1.1")
Returns:
Standard section name (".text", ".rodata", ".data", ".bss") or None
"""
for standard_section, patterns in SECTION_MAPPING.items():
if any(pattern in raw_section for pattern in patterns):
return standard_section
return None
def parse_symbol_line(line: str) -> tuple[str, str, int, str] | None:
"""Parse a single symbol line from objdump output.
Args:
line: Line from objdump -t output
Returns:
Tuple of (section, name, size, address) or None if not a valid symbol.
Format: address l/g w/d F/O section size name
Example: 40084870 l F .iram0.text 00000000 _xt_user_exc
"""
parts = line.split()
if len(parts) < 5:
return None
try:
# Validate and extract address
address = parts[0]
int(address, 16)
except ValueError:
return None
# Look for F (function) or O (object) flag
if "F" not in parts and "O" not in parts:
return None
# Find section, size, and name
for i, part in enumerate(parts):
if not part.startswith("."):
continue
section = map_section_name(part)
if not section:
break
# Need at least size field after section
if i + 1 >= len(parts):
break
try:
size = int(parts[i + 1], 16)
except ValueError:
break
# Need symbol name and non-zero size
if i + 2 >= len(parts) or size == 0:
break
name = " ".join(parts[i + 2 :])
return (section, name, size, address)
return None