On Wed, Sep 7, 2016 at 6:44 AM, Tom Hacohen <t...@osg.samsung.com> wrote: > Hey, > > Good job doing it, but I have to say, I'm quite disappointed you decided > to push it in. > > No tests... Eo is a core piece of infrastructure. I kept it stable and > reliable over the years by making sure every contribution includes > testing, especially big changes like this one. We have no assurances > this change even works correctly (because we don't use threads in the > efl this way yet) and we have no way to check our future changes to see > if they broke it.
well, I assume it passed all existing tests, right. So at least the local thread is working fine. adding more tests for the shared case is indeed required, but not breaking existing code is a good sign. > Shared object and recursive wrapping: I'm not really sure I get what you > were talking about, but if I understand correctly it sounds unnecessary. > People should manage their own mutexes when sharing objects between threads. with his change moving object tables to TLS, doing: ecore_main_loop_begin(); evas_object_move(o, x, y); ecore_main_loop_end(); wouldn't work if "o" was created on the main thread, because "o" is not available in the current TLS. With that, "shared domain" is used to allow this more easily by getting a shared table guarded with locks (what we had before). the alternative would be to use a specific domain, then data_adopt, then execute the method, then data_return. or is that what you mean? remove the "shared" domain and use a single mode for all cases? -- Gustavo Sverzut Barbieri -------------------------------------- Mobile: +55 (16) 99354-9890 ------------------------------------------------------------------------------ _______________________________________________ enlightenment-devel mailing list enlightenment-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/enlightenment-devel
