On 2024/02/15 14:20, Theo Buehler wrote:
> On Thu, Feb 15, 2024 at 12:42:44PM +0000, Klemens Nanni wrote:
> > On Thu, Feb 15, 2024 at 01:37:17PM +0100, Theo Buehler wrote:
> > > Even if you nuke this port, could you try to get a backtrace? SIGILL
> > > might be an IBT issue we need to track down.
> > 
> > Core was generated by `QSyncthingTray'.
> > Program terminated with signal SIGILL, Illegal instruction.
> > #0  0x00000a91ec9cd860 in ?? ()
> > [Current thread is 1 (process 609315)]
> > #0  0x00000a91ec9cd860 in ?? ()
> > #1  0x00000a9135f203df in vmEntryToJavaScript () from /usr/local/lib/libJavaScriptCore.so.0.0
> > #2  0x00000a9135eb6c8e in JSC::JITCode::execute(JSC::VM*, JSC::ProtoCallFrame*) () from /usr/local/lib/libJavaScriptCore.so.0.0
> 
> This does smell a bit like 'NOBTCFI = Yes' could "fix" this...
> 
> If I understood correctly, we need to keep x11/qt5/qtwebkit for the time
> being, so someone with IBT hw should probably hunt this down. This might
> be a starting point:
> 
> https://github.com/WebKit/Webkit/blob/main/Source/JavaScriptCore/llint/LowLevelInterpreter.asm

FWIW ports using it via py-qt5 should be happy, because lang/python has
USE_NOBTCFI=Yes (for this exact reason).

> More importantly, is there a qt6 equivalent of this code that people
> could keep hitting for a longer time?

Qt6 only has qtwebengine (mostly chromium, using a similar set of
patches as www/chromium), no qtwebkit there.