From: Jeff Dike <jdike@ka...> - 2002-07-02 00:40:18
> So we will continue to use uml-devel mailing list as the primary
> discussion group for a while.
Like I said before, that's fine. uml-devel is the right place for this
discussion until it becomes more Windows-related than UML-related.
> Jeff, The next big thing is the task you had outlined a while ago for
> sucking out the platform specific code into platform specific files
> and creating a platform-independent layer. Are you planning on this
> anytime soon?
The timing is more or less up to you. The better UML/Win works, the more
incentive I have to work on this. Also, with more stuff working, the more
material I have to work with wrt seeing what the UML OS interface should
> I can help you out with this. Probably the next patch
> for umlwin32 should be againt the latest uml patch. I noticed that
> kernel text and data has moved into the physical memory region. Is
> there anything else major that I should be aware of?
No, that's the only major structural thing that's happened to UML recently.
And about faulting 16 pages in at a time, I'll have to dredge my memory
to figure out what I was thinking at the time.
From: Jeff Dike <jdike@ka...> - 2002-07-02 01:22:41
> The good news is that Eric Paulson provided me with a way to map pages
> at 4K boundaries ! We are using an obscure undocumented flag for this.
Awesome! That removes the need for that nasty kludge that I can't remember
the details of :-)
Next release of umlwin32 is now available at:
The only big change in this is that I have incorporated Erik Paulson's
hack for mapping at 4K page boundaries. Now we can load any standard
linux binary (including shared libraries). Using hostfs, its now easy to
ftp new programs from a linux box and test to see if they works.
I have tested a few extra commands like vi, objdump, bash with colours
etc. I have also got a C-compiler (http://fabrice.bellard.free.fr/tcc/)
working (check out the /src directory).
Next task: signals.
> The timing is more or less up to you. The better UML/Win works, the more
> incentive I have to work on this. Also, with more stuff working, the more
> material I have to work with wrt seeing what the UML OS interface should
> look like.
I think probably a couple more patches and then we can do this.
> And about faulting 16 pages in at a time, I'll have to dredge my memory
> to figure out what I was thinking at the time.
The good news is that Eric Paulson provided me with a way to map pages at
4K boundaries ! We are using an obscure undocumented flag for this.
I should get this into the next patch which I hope to release before I
take off for July 4th weekend. With this we will be able to run native
binaries, shared library support, command-line editing etc.