Skip to content

BUGFIXES.md

Latest commit

 

History

History
94 lines (63 loc) · 2.78 KB

BUGFIXES.md

File metadata and controls

94 lines (63 loc) · 2.78 KB

Bug Fixes Documentation

This document tracks bugs found and fixed in the Granny dev branch.

Bug #1: FileDirValue validates before creating directory

Issue: FileDirValue.setValue() validates that a directory exists before creating it, causing failure when setting output directory paths.

Location: Granny/Models/Values/FileDirValue.py:28-31

Problem Code:

self.value = value
if not self.validate():
    raise ValueError("Not a directory. Please specify a directory.")
os.makedirs(self.value, exist_ok=True)

Root Cause: The validation happens before directory creation, but validate() checks if directory exists.

Fix: Create directory first, then validate.

Impact: Prevents segmentation analysis from running - crashes during initialization.

Status: FIXED

Fix Applied:

# Create directory first, then validate
os.makedirs(self.value, exist_ok=True)
if not self.validate():
    raise ValueError("Not a directory. Please specify a directory.")

Bug #2: FileNameValue circular validation logic

Issue: FileNameValue.setValue() has flawed logic that tries to validate self.value before it's set to the new value.

Location: Granny/Models/Values/FileNameValue.py:23

Problem Code:

self.value = value if self.validate() else None

Root Cause: validate() checks self.value (old value) instead of the new value parameter.

Fix: Set value first, allow model names that aren't files yet.

Impact: Can cause TypeError when model names are used instead of file paths.

Status: FIXED

Fix Applied:

def setValue(self, value: str):
    self.value = value
    self.is_set = True

def validate(self) -> bool:
    return self.value is not None and (os.path.isfile(self.value) or bool(self.value.strip()))

Testing Strategy

  1. Bug #1 Test: Run segmentation with output directory creation
  2. Bug #2 Test: Use model names vs file paths for segmentation
  3. Integration Test: Full segmentation pipeline with apple tray demo image

Expected result: granny -i cli --analysis segmentation --model pome_fruit-v1_0 --input demo/granny_smith_images/apple_tray should work without manual directory creation.

Test Results

All tests passed!

Integration Test Result:

$ granny -i cli --analysis segmentation --model pome_fruit-v1_0 --input demo/granny_smith_images/apple_tray

	model                    : (user) pome_fruit-v1_0
	input                    : (user) demo/granny_smith_images/apple_tray

0: 704x1024 18 fruits, 25.6ms
Speed: 5.8ms preprocess, 25.6ms inference, 90.8ms postprocess per image at shape (1, 3, 704, 1024)
  • Automatic directory creation works
  • Model name resolution works
  • Segmentation detects 18 fruits successfully
  • No manual directory creation required