Ticket UUID: | 484023 | |||
Title: | small Tcl_MakeFileChannel issue | |||
Type: | Bug | Version: | obsolete: 8.4a4 | |
Submitter: | davygrvy | Created on: | 2001-11-21 03:09:03 | |
Subsystem: | 27. Channel Types | Assigned To: | davygrvy | |
Priority: | 5 Medium | Severity: | ||
Status: | Closed | Last Modified: | 2001-12-07 09:10:09 | |
Resolution: | Invalid | Closed By: | davygrvy | |
Closed on: | 2001-12-07 02:10:09 | |||
Description: |
:Windows If you create a namedpipe which only uses one handle for both read and write and give it to Tcl_MakeFileChannel(), a problem arises when the channel is closed because the close2proc is set to think read and write are 2 different handles and closes each of them. This goes *boom* :) Attached is a patch that I think fixes this. The core does not used namedpipes anywhere, but I found this issue while working on Expect. | |||
User Comments: |
davygrvy added on 2001-12-07 09:10:09:
Logged In: YES user_id=7549 After discovering NamedPipe handles have a notification problem that results in lost data on the other end a pipe connection that in no way is a problem with Tcl's pipe driver (I can see it call WriteFile() just fine!), I conclude this patch is useless :) Who wrote this crappy OS? closing forever... andreas_kupries added on 2001-12-07 04:23:38: Logged In: YES user_id=75003 Given to David so that he sees my question. Just assign it back afterward. andreas_kupries added on 2001-12-07 04:10:55: Logged In: YES user_id=75003 Do we have tests in the testsuite with which we can check that that this change does not break things ? I looked at it and do not believe that it does break anythings, but I would like to check. davygrvy added on 2001-11-21 10:09:04: File Added - 13613: diff.txt |
Attachments:
- diff.txt [download] added by davygrvy on 2001-11-21 10:09:04. [details]