Re: www/palemoon status update

[nia <nia%NetBSD.org@localhost>]

On Fri, Jul 26, 2024 at 07:14:28PM +0100, David Brownlee wrote:
> On Fri, 26 Jul 2024 at 17:44, David Brownlee <abs%absd.org@localhost> wrote:
> >
> > It's great to see more browser options on NetBSD (and in pkgsrc in
> > general) - many thanks for this!
> >
> > Testing a build on NetBSD/amd64-10 from current pkgsrc seems to crash
> > quite often with
> >
> > [19409] ###!!! ABORT: X_CopyArea: BadDrawable (invalid Pixmap or
> > Window parameter); 3 requests ago: file
> > /var/obj/pkg/www/palemoon/work/pale-moon/platform/toolkit/xre/nsX11ErrorHandler.cpp,
> > line 134
> > [19409] ###!!! ABORT: X_CopyArea: BadDrawable (invalid Pixmap or
> > Window parameter); 3 requests ago: file
> > /var/obj/pkg/www/palemoon/work/pale-moon/platform/toolkit/xre/nsX11ErrorHandler.cpp,
> > line 134
> >
> > Running across 'ssh -X localhost' seems to avoid the crash
> > (performance is not really usable, but this was just to get a
> > datapoint)
> >
> > Will try a build with the NS_RUNTIMEABORT commented out to see if
> > everything keeps running as another test
> 
> Adding a check to the NS_RUNTIMEABORT() to call NS_WARNING() for this
> case avoids the crash for me - sample patch below - OK to apply?
> 
> Easy reproducer for me was to try to login to mail.google.com
> 
> Thanks
> 
> David
> 
> $NetBSD$
> 
> Switch NS_RUNTIMEABORT to NS_WARNING to avoid crash seen on NetBSD with
> "X_CopyArea: BadDrawable (invalid Pixmap or Window parameter);"
> 
> --- platform/toolkit/xre/nsX11ErrorHandler.cpp.orig     2024-07-10
> 08:54:53.000000000 +0000
> +++ platform/toolkit/xre/nsX11ErrorHandler.cpp
> @@ -131,7 +131,14 @@ X11Error(Display *display, XErrorEvent *
>  #endif
>  #endif
> 
> -  NS_RUNTIMEABORT(notes.get());
> +  if (message.EqualsLiteral("62")) { // 62 == X_CopyArea
> +    // Avoid NS_RUNTIMEABORT seen on NetBSD with
> +    // "X_CopyArea: BadDrawable (invalid Pixmap or Window parameter);"
> +    NS_WARNING(notes.get());
> +  } else {
> +    NS_RUNTIMEABORT(notes.get());
> +  }
> +
>    return 0; // not reached
>  }
>  }

Yeah, you can commit it. Thanks for testing the browser.

I wonder what other factors could be in play. I run with
layer acceleration forced on (about:config), using i915drmkms.