bug: crash when a requirement is skipped by an environment marker

[vbelouso]

Hi,

I ran into a crash in pybuild-deps when a requirements file contains a dependency that should be ignored on the current platform because of an environment marker.

Environment:

• pybuild-deps: 0.5.0
• Python: 3.14
• OS: Linux

I think this affects non-Windows platforms in general, not just my Linux env.

Minimal repro:

printf "pywin32==311; sys_platform == 'win32'\n" > requirements.txt
pybuild-deps compile -o build.txt requirements.txt

Expected:

On Linux, sys_platform == 'win32' is false, so this requirement should be ignored. My expectation was that the command exits cleanly and produces an empty output file, basically the same behavior as pip / pip-compile.

Actual:

Instead it crashes with:

AttributeError: 'NoneType' object has no attribute 'unsafe_packages'

With --generate-hashes, it crashes in a similar way:

AttributeError: 'NoneType' object has no attribute 'resolve_hashes'

Analysis:

From looking at the code, it seems like inactive marker requirements are still passed further into the resolver. In this case there are no active dependencies to resolve, so the resolver loop does not run and compiler.resolver stays None. Later compile.py assumes the resolver exists and tries to access unsafe_packages or resolve_hashes.

Fix:

The fix I used is:

1. Filter requirements with inactive environment markers in parsers/requirements.py
2. Add defensive handling in scripts/compile.py for the empty dependency case
3. Avoid assuming compiler.resolver is always set

Design note:

One small design point: validation still happens before marker filtering. So this still fails, even on Linux:

echo "foo>1; sys_platform == 'win32'" > requirements.txt
pybuild-deps compile requirements.txt

because `foo>1` is not pinned exactly. I think this behavior is reasonable for this tool, since it catches invalid requirements even if they are only active on another platform.

Refs:

• PEP 508 environment markers
• PyPA dependency specifiers

[gemini-code-assist]

Code Review

This pull request introduces support for filtering out requirements with inactive environment markers during parsing, while still validating all requirements for cross-platform reproducibility. It also updates the compilation script to handle cases where the resolver is null (e.g., when all dependencies are filtered out), and adds corresponding unit and integration tests. Feedback on the tests highlights two main issues: first, using 'sys_platform == 'win32'' as an inactive marker makes tests platform-dependent and prone to failure on Windows, so a guaranteed inactive marker should be used instead; second, using 'traceback.print_tb(result.exc_info[2])' in test assertions can raise a TypeError if 'exc_info' is null and will result in an empty assertion message, so asserting on 'result.exit_code' with 'result.output' is recommended.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

Using sys_platform == 'win32' as an inactive marker makes this test platform-dependent. On Windows, this marker will evaluate to True, causing the assertion assert 'pywin32' not in names and the length assertion to fail. Use a guaranteed inactive marker like sys_platform == 'impossible_platform' to make the test robust across all platforms.

Suggested change

	"pywin32==311 ; sys_platform == 'win32'\n"
	"pywin32==311 ; sys_platform == 'impossible_platform'\n"

[gemini-code-assist]

Using sys_platform == 'win32' as an inactive marker makes this test platform-dependent. On Windows, this marker evaluates to True, so the requirement will be active and the test will fail. Use a guaranteed inactive marker like sys_platform == 'impossible_platform' to ensure the test passes on all operating systems.

Suggested change

	requirements_path.write_text("pywin32==311 ; sys_platform == 'win32'")
	requirements_path.write_text("pywin32==311 ; sys_platform == 'impossible_platform'")

[gemini-code-assist]

Using traceback.print_tb(result.exc_info[2]) in the assertion message has two issues:\n1. If result.exc_info is None (e.g., when the command exits with a non-zero code without raising an unhandled exception), accessing result.exc_info[2] will raise a TypeError, masking the actual assertion failure.\n2. traceback.print_tb prints to stdout and returns None, so the assertion message itself will be empty.\n\nInstead, assert on result.exit_code and provide result.output as the assertion message, which contains the captured stdout/stderr (including any traceback if an exception occurred).

Suggested change

	assert result.exit_code == 0, traceback.print_tb(result.exc_info[2])
	assert result.exit_code == 0, result.output

[vbelouso]

@eskultety can you please take a look?