| Ticket UUID: | 1208108 | |||
| Title: | use Tcl_SaveInterpState | |||
| Type: | Bug | Version: | obsolete: 8.5a3 | |
| Submitter: | dgp | Created on: | 2005-05-24 20:30:15 | |
| Subsystem: | 09. Linked C Variables | Assigned To: | dgp | |
| Priority: | 5 Medium | Severity: | ||
| Status: | Closed | Last Modified: | 2005-07-06 22:11:08 | |
| Resolution: | Accepted | Closed By: | dgp | |
| Closed on: | 2005-07-06 15:11:08 | |||
| Description: |
When LinkTraceProc handles write traces, it calls on various Tcl_GetFooFromObj() routines to check the "type" of the written value, against the link type. The "interp" is passed to these calls to capture error messages. When there's no error, the impact on the interp isn't supposed to be perceived, and LinkTraceProc saves and restores the interp result to achieve this. The interp result is only a part of the interpreter state. The new routines Tcl_(Save|Restore)InterpState would be better tools to completely achieve this task. | |||
| User Comments: |
dgp added on 2005-07-06 22:11:06:
Logged In: YES user_id=80530 Patch applied to HEAD. *not* safe for 8.4. dgp added on 2005-07-06 21:56:06: File Added - 141055: 1208108.patch dgp added on 2005-07-06 21:56:04: Logged In: YES user_id=80530 With that in mind, the attached patch simplifies LinkTraceProc a good deal. dgp added on 2005-07-06 21:44:39: Logged In: YES user_id=80530 I think I misread this one. LinkTraceProc() is a Tcl_VarTraceProc. It gets called only by TclCallVarTraces, which does its own save/restore of the interp state, precisely so each trace routine doesn't have to. dkf added on 2005-05-25 16:16:45: Logged In: YES user_id=79902 Sounds fair to me | |||
Attachments:
- 1208108.patch [download] added by dgp on 2005-07-06 21:56:05. [details]