| Summary: | Race in TAO transport | ||
|---|---|---|---|
| Product: | TAO | Reporter: | Nanbor Wang <bala> |
| Component: | ORB | Assignee: | DOC Center Support List (internal) <tao-support> |
| Status: | NEW --- | ||
| Severity: | normal | CC: | bala_natarajan, duane_binder; |
| Priority: | P3 | ||
| Version: | 1.5.1 | ||
| Hardware: | All | ||
| OS: | All | ||
Bala, can you add the test case to this entry? Duane, can you had your test case to this bugzilla entry? |
The race comes to wreak havoc only when wchar encodings are used between the client and the server along with corbalocs. Duane Binder at Symantec has a good test that reproduces this consistently. I will get his test case and attach to this bugzilla. Here is the race ------------------- TAO_InputCDR input_cdr (db, flg, rd_pos, wr_pos, qd->byte_order_, qd->major_version_, qd->minor_version_, this->orb_core_); transport->assign_translators(&input_cdr,&output); ----------------------------------- assign_translators () can be called by two different threads at the same time. Despite charectar set encoding rules, I can think of cases with corbalocs where the resets can happen. To cut the long story short, there is a race. But the problem is a lot complex. I am just not providing a solution for now since it could be wrong.