fix(dir): use Prefix instead of StartAfter for directory existence check

[alexsavio]

Pull Request: Use Prefix instead of StartAfter for directory existence check

Summary

This PR replaces the StartAfter parameter with invalid UTF-8 sequences with a simpler Prefix-based query when checking for directory existence in intelligentListCut.

Problem

The current implementation in core/dir.go uses StartAfter with a crafted UTF-8 suffix (.\xEF\xBF\xBD) to simulate a >= operator for directory existence checks:

StartAfter: PString(lastName[0:lastLtPos] + ".\xEF\xBF\xBD"),

This approach:

1. Relies on specific UTF-8 byte ordering behavior
2. Uses invalid/replacement UTF-8 characters that may be handled differently by various S3-compatible backends
3. Is complex and hard to understand (requires understanding UTF-8 code point ranges)

We encountered issues with this approach when using AWS S3 in our setup.

Solution

Replace the StartAfter approach with a direct Prefix query:

Prefix: PString(lastName[0:lastLtPos] + "/"),

This is:

• More straightforward: Directly queries for items starting with path/
• More portable: Works consistently across different S3-compatible backends
• Easier to understand: No need to understand UTF-8 byte ordering tricks

Changes

1. core/dir.go: Replace StartAfter with Prefix in intelligentListCut
2. core/dir_test.go: Update test assertion to match new behavior
3. core/goofys_common_test.go: Add default CLOUD=s3 for easier local test execution

Testing

• Updated existing unit tests to verify the new behavior
• Tested with AWS S3 backend

Related

This is similar in spirit to commit de62549 which removed StartAfter-based ext-v1 autodetection because it didn't work with Yandex S3.

---

How to apply these changes

If you want to create this PR from your own fork:

# Add upstream remote if not already added
git remote add upstream git@github.com:yandex-cloud/geesefs.git

# Fetch latest upstream
git fetch upstream

# Create a new branch from upstream master
git checkout -b fix-intelligent-list-cut-utf8 upstream/master

# Apply the patch (see PATCH.diff file) or manually make the changes
git apply PATCH.diff

# Commit
git commit -m "fix(dir): use Prefix instead of StartAfter for directory existence check"

# Push to your fork
git push origin fix-intelligent-list-cut-utf8

# Then create PR via GitHub UI