| Ticket UUID: | 1969953 | |||
| Title: | Buffersize should clip, not ignore, big values | |||
| Type: | Bug | Version: | None | |
| Submitter: | ferrieux | Created on: | 2008-05-22 21:57:07 | |
| Subsystem: | 24. Channel Commands | Assigned To: | andreas_kupries | |
| Priority: | 5 Medium | Severity: | ||
| Status: | Closed | Last Modified: | 2008-05-24 04:34:45 | |
| Resolution: | Accepted | Closed By: | andreas_kupries | |
| Closed on: | 2008-05-23 21:34:45 | |||
| Description: |
The [fconfigure -buffersize] value is documented as having a valid range of [1,1000000], fine. First, this is inaccurate, since the code uses the value 1M (1024*1024, not 1000*1000) instead. Second, when a value larger than 1M is given, it is silently ignored, staying at its previous value. In most cases this means the default of 4k. Bottom line: people try to improve their throughput on large transfers by setting large values, and it has no effect, and no error message tells them about the mechanism. I find this behavior stupid. We should: - either clip large values to the max allowed - or raise an error if out of range Please find attached a patch doing the former. | |||
| User Comments: |
andreas_kupries added on 2008-05-24 04:34:45:
Logged In: YES user_id=75003 Originator: NO Committed to Tcl 8.4 branch. Updated testsuite. Committed to Tcl 8.5 branch. Updated testsuite. Committed to Tcl Head branch. Updated testsuite. ferrieux added on 2008-05-23 05:02:59: Logged In: YES user_id=496139 Originator: YES To argue further for doing something, see the number of raised reds herrings due to this behavior. Thinks about the total number of lost man-hours over the (unknown) number of silent sufferers..?? http://groups.google.com/group/comp.lang.tcl/browse_frm/thread/648eb6bfff8da689/f99c56badea40470#f99c56badea40470 ferrieux added on 2008-05-23 04:57:09: File Added - 278729: buffersize.patch | |||
Attachments:
- buffersize.patch [download] added by ferrieux on 2008-05-23 04:57:08. [details]