Would someone please fill me in on direction for loader code?

[Jefferson Carpenter]

When I write patches touching the loader code (including process and 
thread initialization and startup), it would be useful to know the 
overall plans for that area.  Generally I'm asking about ntdll.

What is the team's overall direction for that area of code?  Is it as 
simple as "it just has to work"?  Are there things about how it works 
that aren't expressed in code but have to be preserved?  Or do you have 
plans for how you're intending to change it in the future?


Thanks,
Jefferson

[Stefan Dösinger]

> Am 02.09.2020 um 13:29 schrieb Jefferson Carpenter <jeffersoncarpenter2(a)gmail.com>:
>
> What is the team's overall direction for that area of code?  Is it as simple as "it just has to work"?  Are there things about how it works that aren't expressed in code but have to be preserved?  Or do you have plans for how you're intending to change it in the future?

It is complicated and always evolving. The ntdll tests certainly capture a lot of requirements, but not everything. What's making matters more complicated is that DRM and anti-cheat systems hook into ever-lower levels of ntdll, and we have to do our best to replicate how things work on Windows.

What is it that you are trying to do in the loader code?

[Jefferson Carpenter]

On 9/2/2020 11:04 AM, Stefan Dösinger wrote:
> > Am 02.09.2020 um 13:29 schrieb Jefferson Carpenter <jeffersoncarpenter2(a)gmail.com>:
> >
> > What is the team's overall direction for that area of code?  Is it as simple as "it just has to work"?  Are there things about how it works that aren't expressed in code but have to be preserved?  Or do you have plans for how you're intending to change it in the future?
>
> It is complicated and always evolving. The ntdll tests certainly capture a lot of requirements, but not everything. What's making matters more complicated is that DRM and anti-cheat systems hook into ever-lower levels of ntdll, and we have to do our best to replicate how things work on Windows.
>
> What is it that you are trying to do in the loader code?

Well, right now I have bugs 49755, 49754, and 49495 open indicating 
valgrind warnings in the ntdll code.  I recently submitted "[PATCH] 
ntdll: Decouple pthread stack and win32 stack." to try to fix the last 
one, but it was rejected for good reasons.

It's good to know you have pretty thorough test coverage.

Those bugs might be closed without needing to be patched if the warnings 
are all false positives.  However as I look into matters I might come 
across some useful patches.  I just want to submit worthwhile patches 
that won't break anything.

-Jefferson