| Ticket UUID: | 293344d4f3e07f7f9da93db4b967abed8122bd42 | ||
| Title: | Regression in SQLite test-suite | ||
| Type: | Bug | Created on: | 2017-07-13 15:00:42 |
| Submitter: | jan.nijtmans | Assigned to: | nobody |
| Subsystem: | - New Builtin Commands | Severity: | Minor |
| Priority: | 5 Medium | Last modified: | 2017-07-13 15:08:55 |
| Status: | Closed | Closed by: | jan.nijtmans |
| Resolution: | Fixed | Closed on: | 2017-07-13 15:08:55 |
| Version: | core-8-6-branch | ||
| Description: | ||||
|
When running the latest SQLite test-suite there was a regression in 2 test-cases, one of them: badutf2-4.1.5... ! badutf2-4.1.5 expected: [\u0080] ! badutf2-4.1.5 got: [\uFF80] bisecting Tcl pointed to commit [7351fc0c9c], but - actually - there's nothing wrong with this commit. Further investigation resulted in the observation that the macro TclUtfToUniChar() contained a mistake: If 'str' points to a signed character (which it normally does), a character in the range \x80-\xC0 results in an unicode character in the range \xFF80-\xFFC0 ... this is clearly wrong. Conclusion: this is a long-standing bug in the TclUtfToUniChar() macro, which was never discovered because there was no test-case touching this. Commit [7351fc0c9c] merely exposed this bug in the SQLite test-suite. | ||||
| User Comments: | ||||
jan.nijtmans added on 2017-07-13 15:08:55:
Fixed in core-8-5-branch, core-8-6-branch and trunk | ||||