Re:The cats sitting on a fence in early builds of Windows 8
by laumars on Wednesday February 23, 2022 @03:03AM
(#30438783)
Microsoft likely didn’t create that art. So I suspect a bigger reason for removing it was licensing, with “optimisation” being a strong excuse for not spending resource on creating original assets.
Re:The cats sitting on a fence in early builds of Windows 8
by Brian_K_White on Wednesday February 23, 2022 @06:52AM
(#30440105)
Joan G Stark
Re:The cats sitting on a fence in early builds of Windows 8
by fredoralive on Wednesday February 23, 2022 @06:46AM
(#30440071)
The kernel text mode font is probably baked into the kernel and not accessible from user space / Win32?
Re:The cats sitting on a fence in early builds of Windows 8
by exikyut on Wednesday February 23, 2022 @02:20AM
(#30438566)
I *think* this was part of the shell layer. Which is why it's fascinating how the performance team wanted to prune it out; presumably the fonts were mapped into memory elsewhere and thus deduplic...oh, maybe Windows doesn't have mmap page dedupe.
Re:The cats sitting on a fence in early builds of Windows 8
by monocasa on Wednesday February 23, 2022 @03:12AM
(#30438835)
NT's VFS cache does a decent enough job at sharing pages to actual data for the same file, even between kernel space and user space; the problems start around FS metadata caching.
Re:The cats sitting on a fence in early builds of Windows 8
by raverbashing on Wednesday February 23, 2022 @02:52AM
(#30438739)
Edit: I don't think the kernel would render TTF and/or use the full graphical stak so that's probably the main difference
It probably has, but maybe not between kernel and userspace
Re:The cats sitting on a fence in early builds of Windows 8
by formerly_proven on Wednesday February 23, 2022 @03:33AM
(#30438994)
> Edit: I don't think the kernel would render TTF and/or use the full graphical stak so that's probably the main difference
Oh but it does. Most (all?) text rendering on Windows happens in the kernel.
Re:The cats sitting on a fence in early builds of Windows 8
by fredoralive on Wednesday February 23, 2022 @06:56AM
(#30440130)
AFAIK stuff like .net's WPF and the GPU accelerated DirectWrite do font rendering in userspace. It's ye olde GDI that is in the kernel, because NT didn't run well on a 486[1]. Yes, Windows has about 5 different ways to render fonts, and no, they don't seem to render fonts the same way.
(In the case of this kernel text mode / BSOD code, I do wonder does it use the GDI font rendering, or does it have its own separate mini renderer just for those screens?)
[1] Probably a bit of a simplification, I know.
Re:The cats sitting on a fence in early builds of Windows 8
by formerly_proven on Wednesday February 23, 2022 @10:43AM
(#30442248)
I don't know anything about DirectWrite, but there seems to be a DirectWrite kernel component due to CVE-2020-1577 (DirectWrite leaking kernel memory "by convincing a user to visit an untrusted webpage", which I'm not sure whether the exploit route would be in TTF/OTF processing, e.g. due to webfonts, or the text rendering itself).
Re:The cats sitting on a fence in early builds of Windows 8
by raverbashing on Wednesday February 23, 2022 @03:36AM
(#30439016)
I stand corrected then! (Unless this changed recently)
https://googleprojectzero.blogspot.com/2016/06/a-year-of-win...
Re:The cats sitting on a fence in early builds of Windows 8
by zinekeller on Wednesday February 23, 2022 @07:37AM
(#30440348)
I mean, in recent version of Windows it's now on the user-space, but because Microsoft retains bug-for-bug compatibility Windows 8 will be stuck with a kernel font rasteriser*
* DirectWrite and WPF have been always user-space.
But iirc, Windows 8 introduced Segoe Mono and the kernel text mode interface is/was using that font, too?