Ah that explains a lot of things, thanks for walking me through this! On Friday, August 9, 2024 at 8:55:31 PM UTC+9 omer...@chromium.org wrote:
> Thanks, but note that your last test is not guaranteed to always succeed. > You're checking that some value is not found on stack, and trying to > prevent leftover stale values on stack to make the test pass. > Most times this will likely work, but you can't 100% guarantee that some > value will not be present on the stack, and if it does your test will > flakily fail. > This could happen if e.g. a pointer somehow leaks to stack unexpectedly > (stale value, some compiler optimization, etc...) or just a random false > positive (i.e. the address of your object randomly appearing on stack > regardless of any actual pointers to it). > > On Friday, August 9, 2024 at 1:53:15 AM UTC+2 zcb...@gmail.com wrote: > >> For future reference, here is the test that verifies pointers in >> std::variant can be retained but pointers in std::vector can not: >> >> https://github.com/compilets/compilets/blob/abb49a4/cpp/runtime/tests/stack_unittest.cc >> >> On Thursday, August 8, 2024 at 9:59:11 PM UTC+9 Cheng wrote: >> >>> Thanks for the explanation! >>> >>> I was actually quite confused why pointer in vector was also retained, >>> I'll fix my tests. >>> >>> On Thursday, August 8, 2024 at 8:00:44 PM UTC+9 omer...@chromium.org >>> wrote: >>> >>>> For some reasons my previous replies didn't get to the mailing list. >>>> >>>> The question is whether the pointer itself is on the stack or not. >>>> IIRC std::variant uses inline storage to store values, so a pointer >>>> that you keep in a std::variant would be on stack and would be found by >>>> stack scanning. >>>> std::vector, on the other hand, allocates an off-stack backing store >>>> that it allocates (and reallocates as the vector grows), so pointers kept >>>> in a std::vector would not be found by stack scanning. >>>> >>>> If your tests pass, it's because the GC is finding your pointer >>>> somewhere else on the stack (e.g. left over from calling set_needle), but >>>> not in the vector. >>>> On Thursday, August 8, 2024 at 10:43:43 AM UTC+2 zcb...@gmail.com >>>> wrote: >>>> >>>>> To answer my own question, I wrote some tests: >>>>> >>>>> https://github.com/compilets/compilets/commit/d69722fb36260caa278843311cb043f5fe8a90d7 >>>>> >>>>> It seems that pointers stored in containers can be retained. >>>>> (It would be great if someone can verify this though.) >>>>> >>>>> On Thursday, August 8, 2024 at 9:11:33 AM UTC+9 Cheng wrote: >>>>> >>>>>> Cppgc does stack scanning so objects pointed by raw pointers on stack >>>>>> are not garbage collected: >>>>>> >>>>>> Object* ptr = MakeGarbageCollected<Object>(); // retained. >>>>>> >>>>>> But what if I put the pointer in a variant? >>>>>> >>>>>> std::variant<Object*, std::monostate> ptr = >>>>>> MakeGarbageCollected<Object>(); // retain? >>>>>> >>>>>> Or even in a vector? >>>>>> >>>>>> std::vector<Object*> ptrs = { MakeGarbageCollected<Object>() }; // >>>>>> retain? >>>>>> >>>>>> Will the object still be retained by the container of pointer on >>>>>> stack? >>>>>> >>>>> -- -- v8-users mailing list v8-users@googlegroups.com http://groups.google.com/group/v8-users --- You received this message because you are subscribed to the Google Groups "v8-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to v8-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/v8-users/286a00cb-873a-4c60-be87-cc3fbeed125fn%40googlegroups.com.
