| Ticket UUID: | bc7925fc5038e49d114fb6bc0d6d8366f9be2f67 | ||
| Title: | Inconsistencies in long path handling in Tcl 8.6 and 9 | ||
| Type: | Bug | Created on: | 2025-12-18 11:10:06 |
| Submitter: | apnadkarni | Assigned to: | nobody |
| Subsystem: | - New Builtin Commands | Severity: | Minor |
| Priority: | 5 Medium | Last modified: | 2025-12-18 11:15:49 |
| Status: | Open | Closed by: | nobody |
| Resolution: | None | Closed on: | |
| Version: | |||
| Description: | ||||
|
(I'm working on adding support for long paths as available in newer Windows versions. The bugs below are only tangentially related as the new support does not use the Tcl 9 (and 8.6) support for long file paths (> MAX_PATH characters) on Windows is inconsistent and unusable from applications. Consider
All that works, but leads to inconsistent behaviour like you can create the directory but not change to it
Or the directory exists but not its parent!
Or discrepancies with commands like
The first case should not be returning an error. The second case should be returning the leaf component. The basic issue is that Tcl tries to handle Windows
The result is the inconsistent and confusing behavior as logged in the c.l.t. thread and Tk ticket. My opinion is that Tcl should not try to fix up deficiencies in a system API (Win32 in this case) unless it can do a reasonably thorough job. It is better to raise an error than present a logically inconsistent interface to the user as illustrated in the above commands. | ||||
| User Comments: | ||||
apnadkarni added on 2025-12-18 11:15:49:
As a point of clarification, the In particular, it will not help on systems where long path support has not been enabled. On these systems the issues listed in the ticket will continue to exist. | ||||
