tommie
tommie
> Not a TMC expert, just to mention there is no buffering inside `usbd_edpt_xfer()`, you need to add buffer like how bulk endpoints did: > > https://github.com/hathach/tinyusb/blob/9e674fa109348d97fb4280ac534be665f619a4d8/src/class/usbtmc/usbtmc_device.c#L129 Yes, I documented...
Is there a regression test for this? In what circumstances is it a problem? Since `MarkSlotsAllocated` does it this way, this change makes sense to me.
(It would be nice if this bot provided WAT, like the one in #465.) I think this is because the recursive call in `CopyStackSlotsR` assumes `numSlots` is 1, which fails...
Here is a minimal test case: ```wat (module (func (;0;) block $1 (result f64 f64) block (result f64 f64) f64.const -0x1.ba17b2943f09fp-37 f64.const -0x1.ba17b2943f09fp-37 br $1 end end drop drop) (export...
@vshymanskyy said on #489 > This change has been reverted as it breaks many CI checks Sorry about that! I had a look at the failures in https://github.com/wasm3/wasm3/actions/runs/10222976517 They all...
It seems #491 may have solved this issue too. The minimal test case passes without ASAN issues already with #491.
This seems to have something to do with branching out of a block where the block result is declared as fewer than what's on the stack. Minimal test case: ```wat...
This is an amusing one... I can't make it smaller than https://gist.github.com/tommie/90c3ec632d24378d08630cbc2e215abf Valgrind finds the same issue in release mode. Notice the "WOW" comment. If you change that constant (which...
That isn't the full story. If I replace `f64` with `i64`, even the ~~correct~~ eight `locals` will see ASAN fail, and even with WOW being zero. Adding a ninth local...
Sorry, I confused myself about locals and stack depth. The correct number of locals is zero here. That makes it even more amusing that adding locals solves issues.