undroidwish

This is the patch for tclXprofile.c, no Tcl core changes seem necessary https://www.androwish.org/index.html/fdiff?v1=a6f5cf9f8b21bf4d&v2=eaeaabfa2e4ac909&patch

Stupid me, I was on a wrong old TclX checkout. Seems some of the work was done 2 years ago. Following the ticket http://core.tcl.tk/tcl/tktview/cd82cec7ce46a55af099 I now understand the problem with...

By Tcl_Obj'fying parts of TraceExecutionProc() of tclTrace.c the overhead can be reduced to about 60% which is still slow compared to unprofiled execution, see attached patch. But this is a...

Here's the slightly improved patch: [tclTrace.diff.txt](https://github.com/flightaware/Tcl-bounties/files/666713/tclTrace.diff.txt) And this script (including some recorded output) was used to get an idea on profile's timing impact: [perftest.tcl.txt](https://github.com/flightaware/Tcl-bounties/files/666721/perftest.tcl.txt) Conclusion: profiled execution is 100 to...

Regarding the "yield" effect/crash please would you post a test case? I've tried this snippet (w/o the yield patch) and had no crashes (on 32bit i386 with core-8-6-branch plus Tcl_Obj'fication):...

Apparently, the first version of Tcl_Obj'fication wasn't a big hit. Here's the second version which uses an array of parsed Tcl_Obj's attached to the TraceCommandInfo struct: [tclTrace-v2.diff.txt](https://github.com/flightaware/Tcl-bounties/files/674182/tclTrace-v2.diff.txt)

Turns out that the latest patch doesn't honour this important sentence from "man Tcl_EvalObjv": "Tcl_EvalObjv executes a **single** pre-parsed command instead of a script." Thus, the new strategy is to...

My advise in your case is to go the multi threaded path, i.e. have a dedicated thread serving OPC/UA and let the rest of your framework perform its usual tasks....