What

_log_pretty_url is documented to "remove the protocol and www. prefix", but it implemented this with str.replace:

s = s.replace('https://', '').replace('http://', '').replace('www.', '')

str.replace removes every occurrence, not just the leading prefix, so the helper corrupts otherwise valid URLs:

• https://awwww.com → awcom (the substring www. inside the domain is stripped)
• https://example.com/r?u=https://other.com → example.com/r?u=other.com (embedded scheme in the query is dropped)
• https://site.com/www.assets/img → site.com/assets/img (www. inside the path is stripped)

This shows up in the tab-info debug logging in browser/session.py, making logged URLs misleading.

Fix

Strip only a leading scheme and a single leading www., matching the documented behavior:

s = re.sub(r'^https?://', '', s)
if s.startswith('www.'):
    s = s[len('www.') :]

The rest of the URL (path, query, fragment) is left untouched.

Verification

Added tests/ci/infrastructure/test_log_pretty_url.py covering prefix stripping, the three corruption cases above, and the existing truncation behavior.

$ uv run pytest tests/ci/infrastructure/test_log_pretty_url.py -p no:xdist -o addopts="" -q
5 passed

Before the fix the three "does not corrupt" cases fail; after the fix all 5 pass. ruff check and ruff format --check pass on the changed files.

Note: drafted with AI assistance and verified by a human (RED/GREEN tests run locally).

Summary by cubic

Fix _log_pretty_url to only strip a leading http(s):// and a single leading www., preventing mangled domains and embedded URLs in logs. Adds focused tests; truncation behavior is unchanged.

• Bug Fixes
  • Replaced global str.replace with re.sub(r'^https?://', ...) and a prefix check for www..
  • Preserves path/query and embedded https:///www..
  • Added tests in tests/ci/infrastructure/test_log_pretty_url.py.

^{Written for commit aaa6033. Summary will update on new commits.}