Ticket UUID: | 67159f33bbd7e3d44aa0dba13e463340970a13e5 | |||
Title: | epoll notifier and --disable-threads build | |||
Type: | Bug | Version: | 8.7 | |
Submitter: | dgp | Created on: | 2017-08-10 14:15:27 | |
Subsystem: | 01. Notifier | Assigned To: | jan.nijtmans | |
Priority: | 5 Medium | Severity: | Critical | |
Status: | Closed | Last Modified: | 2018-05-18 08:31:49 | |
Resolution: | Wont Fix | Closed By: | jan.nijtmans | |
Closed on: | 2018-05-18 08:31:49 | |||
Description: |
As implemented, the new epoll notifier makes direct calls to the pthread_*() system routines. These routines are not available in a --disable-threads build, and the notifier fails. Either we need appropriate build safeguards to prevent building the epoll notifier in this configuration, or perhaps the epoll notifier might be re-implemented making calls to the Tcl_* threading primitives which might possibly do the right thing in that configuration. | |||
User Comments: |
jan.nijtmans added on 2018-05-18 08:31:49:
Since TIP #491 is accepted now, --disable-threads builds will be discontinued anyway. Therefore we don't have to fix this any more. jan.nijtmans added on 2017-08-17 22:59:57: My suggestion would be: If Tcl is built with --disable-threads, then the epoll notifyer et al could be disabled as well. Since Epoll depends on the pthread functions, I wouldn't want to change the epoll implementation for this. I'll see if I can give a try on this. ferrieux added on 2017-08-10 18:06:28: Yes. The gist of that conversation being whether, in this new notifier, being threaded implies a dedicated extra thread (hence process table entry) or not. If not, the main reason for keeping --disable-threads alive vanishes. dgp added on 2017-08-10 14:20:07: I suppose another possible answer is to end support for --disable-threads building, but I think we'd want to have a careful conversation about that. |