| Ticket UUID: | 1284178 | |||
| Title: | [format %h?] rejects valid integers | |||
| Type: | Bug | Version: | obsolete: 8.4.11 | |
| Submitter: | dgp | Created on: | 2005-09-07 18:17:48 | |
| Subsystem: | 16. Commands A-H | Assigned To: | dkf | |
| Priority: | 5 Medium | Severity: | ||
| Status: | Closed | Last Modified: | 2005-10-14 04:46:33 | |
| Resolution: | Fixed | Closed By: | dkf | |
| Closed on: | 2005-10-13 21:46:33 | |||
| Description: |
% format %hu 0xffff 65535 % format %hu 0x10000 0 % format %hu 0xffffffff 65535 % format %hu 0x100000000 integer value too large to represent It's clear that %hu truncates to short int size before formatting, but fails to do the same truncation when given a wide integer value. It just rejects those, without a good reason. | |||
| User Comments: |
dkf added on 2005-10-14 04:46:33:
Logged In: YES user_id=79902 Fixed in 8.4 branch. dgp added on 2005-10-14 03:40:35: Logged In: YES user_id=80530 my position is that the tests themselves are broken. dkf added on 2005-10-14 03:27:51: Logged In: YES user_id=79902 Fixing this breaks format-17.1 and format-18.2 dgp added on 2005-10-14 00:52:02: Logged In: YES user_id=80530 fixed on HEAD by TIP 237 merge. Still needs attached patch applied to fix in 8.4.11 dgp added on 2005-09-12 08:54:31: File Deleted - 148370: File Added - 148760: 1284178.patch dgp added on 2005-09-12 08:54:29: Logged In: YES user_id=80530 corrected patch now attached dgp added on 2005-09-11 08:58:21: Logged In: YES user_id=80530 Do not accept that patch. It introduces a new bug: % format %ld [expr wide(3000000000)] -1294967296 dgp added on 2005-09-08 01:34:05: File Added - 148370: 1284178.patch dgp added on 2005-09-08 01:34:03: Logged In: YES user_id=80530 Here's a patch that makes all of [format]'s integer conversion specifiers accept all of Tcl's valid integer values. dgp added on 2005-09-08 01:29:50: Logged In: YES user_id=80530 A related issue is integer [format]s without a length specifier, for example: % format %u 0xffffffff 4294967295 % format %u 0x100000000 integer value too large to represent As with the "h"-constrained formatting, we could truncate to the native length of a long before formatting and accept the wide argument. Doing that would produce: % format %hu 0x100000000 0 However, this behavior is actually tested by the test suite via tests format-17.1 and format-18.2 . The restriction and these test don't really make sense to me, but perhaps a maintainer knows better? | |||
Attachments:
- 1284178.patch [download] added by dgp on 2005-09-12 08:54:29. [details]