cgsqlite,.: fix regression with empty strings

[raggi]

A final step refactor in 81bdfd0 introduced a bug where empty strings became NULL in some code paths.

This fixes the regression and introduces a test to prevent reoccurrence.

Updates tailscale/corp#9199

[Copilot]

Pull request overview

This PR fixes a regression where empty strings were incorrectly treated as NULL values when binding parameters to SQLite statements. The issue was caused by unsafe.StringData("") potentially returning nil, which SQLite interpreted as NULL.

Key Changes:

• Fixed the empty string handling in cgosqlite/cgosqlite.go by using a backing array to guarantee a non-nil pointer
• Added comprehensive regression test TestEmptyStringNotNull to verify empty strings are properly distinguished from NULL values

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 1 comment.

File	Description
cgosqlite/cgosqlite.go	Replaced unsafe.StringData("") with a backing array approach to ensure empty strings have a non-nil pointer representation
sqlite_test.go	Added comprehensive test verifying empty strings are stored and retrieved correctly, distinct from NULL values

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[bradfitz]

What was the point of that unreviewed change anyway? What was wrong with bind_text64_empty?

[raggi]

What was the point of that unreviewed change anyway? What was wrong with bind_text64_empty?

I noticed the comment over the functions was wrong, and realized they no longer had a good reason to be. It should have been a no-op, but the stringdata behavior caught me, and that's my bad.