Ticket UUID: | 721ec6927136c79169195ac09cfe6efd0a27fce6 | |||
Title: | stacked channel self-manipulation protections too weak | |||
Type: | Bug | Version: | 8.5 + | |
Submitter: | dgp | Created on: | 2014-03-27 16:25:08 | |
Subsystem: | 25. Channel System | Assigned To: | dgp | |
Priority: | 9 Immediate | Severity: | Severe | |
Status: | Closed | Last Modified: | 2014-05-08 17:42:48 | |
Resolution: | Fixed | Closed By: | dgp | |
Closed on: | 2014-05-08 17:42:48 | |||
Description: |
When the read operation on a channel can cause stack/unstack operations to re-shape the structure of the channel, we can get into a situation where pointers point to bad places. There are some protections from this in the code, but not enough. Committing new test iortrans-4.8.1 to trunk to expose such a segfault. | |||
User Comments: |
dgp added on 2014-04-24 17:29:55:
The segfaults are squashed now. dgp added on 2014-04-22 15:10:17: iogt-2.4 segfaults fixed with addition of refcounting machinery and preservations to the [testchannel transform] implementation. Test still fail. Have to ponder what proper behavior is here. dgp added on 2014-04-21 19:18:12: The latest commit to trunk (refCounting of ChannelBuffer) is sufficient to stop the segfault in test iortrans-4.8.1. Now it's just a failing test. iogt-2.4 still segfaults. dgp added on 2014-03-27 16:44:47: Test iogt-2.4 shows it is an 8.5 issue too. |