|
2008-04-01
| ||
| 05:06 | • Ticket [1805794fff] changing unix system time status still Open with 4 other... artifact: bbd2ee8ff1 user: ferrieux | |
|
2007-10-01
| ||
| 16:54 | • New ticket [1805794fff]. artifact: 11a257d216 user: kitchen | |
| Ticket UUID: | 1805794 | ||
| Title: | changing unix system time | ||
| Type: | Bug | Created on: | 2007-10-01 16:54:21 |
| Submitter: | kitchen | Assigned to: | kennykb |
| Subsystem: | 03. Timer Events | Severity: | |
| Priority: | 5 Medium | Last modified: | 2008-04-01 05:06:19 |
| Status: | Open | Closed by: | |
| Resolution: | None | Closed on: | |
| Version: | None | ||
| Description: | ||||
TCL is unable to deal with Unix time being reset to a value less than the current time. It causes timers to stop working until the time "catches up" to where it was originally set. For example if I set the system time backwards by 100 seconds and I have programmed the system for a 100mS timer I will not receive a timer event for 100 seconds. In my case I happen to be doing this from the C language interface using Tcl_CreateTimerHandler() however I'm sure this affects all aspects of timer operation under TCL. Setting the system time forward is less of a problem. It simply causes the timers to immediately wake up -- assuming that the time has been set forward far enough. I realize this may be difficult to solve. select() doesn't have a problem with it. I suspect that select() must use the kernel jiffy counter rather than looking at the system time. A trick I've used elsewhere in my code is to look at the system time when select() wakes up and compare it to the previous pass and if the difference is several times larger than the select() sleep time assume the clock has been reset and adjust things in the program accordingly. | ||||
| User Comments: | ||||
ferrieux added on 2008-04-01 05:06:19:
Logged In: YES user_id=496139 Originator: NO See also TIP 302 for an analysis of the possible solutions. http://www.tcl.tk/cgi-bin/tct/tip/302.html | ||||
