[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: Hello all

hi to everybody

quite a nice discussion we are having

my 2 cents (of euro) 

On Thu, Nov 15, 2001 at 09:58:59AM +1100, Robert Collins wrote:
> Some of you here may recognise me - among other things I have just taken
> on maintenance of cygwin's setup.exe, and have been a cygwin net
> contributor for a while now. (Chris - how long exactly ?).

wow thanks for joining

> 1) Replacing open files. Say that setup uses berkley db3 as a .dll.
> Setup can not replace that .dll itself - and any dpkg/rpm style port
> will assume that it can replace that .dll. I've made a beta release of
> setup.exe that *can* do this, but it needs testing and work.

disclaimer: I have never used cygwin inside windows; my experience
with cygwin has been: I have cross compiled 'gtkmorph' , compiling
it in Debian/GNU/Linux to work in win32 (and it works, although buggy);
so I am not sure of what I am saying

so here is a quick idea for the above: is it possible to rename an open file?
if so, dpkg will just rename the files before installing a new copy
of the same file from a .deb
(some other program will then delete the old file when it is not in use
anymore: you name a way...)

> 2) Porting issues. Setup.exe is a win32 program, not a cygwin program.
> It could dynamically open cygwin1.dll, once it's been installed, but the
> core functionality, to grab packages off the net and install cygwin1.dll
> in the correct place with the correct permissions has to be built in.
> Thus any dpkg port that will bootstrap debian-w32 needs to be a native
> win32 program, statically linked to any libraries, and able to bootstrap
> itself to get any required .dll's.
> I think it would be a shame for users to have to bootstrap cygwin from
> cygwin.com, and then grab a *different* installer for debian-w32. IMO
> setup.exe should be a direct bootstrap. Also I think that maintaining a
> separate tree of binaries and source does not make sense for Cygwin
> today. There simply are not enough kernel developers or package
> maintainers at this point. It would be great to see some of the debian
> features and capabilities brought to the existing environment IMO.

my vision of Debian/w32 is different

in the future, when Debian/w32 will be a reality, there will be a file
called Deboostrap.exe in the debian archive; the user will download this
file and execute it: this file will ask where the user wishes to install
Debian/w32, and then will selfextract from inside (more or less 
as the selfextracting archives do) a  version base.tgz (Debian people
know what it is) inside the above-asked directory; it will then
setup virtual mounts, and proceed with normal Debian installation

idem est: Deboostrap.exe will substitute the Debian boot diskettes,
and the Debian/w32 installation will be the same as Debian/i386
after that

Deboostrap.exe may even be just  a dos program
(in that case, it will ask 
'where do you want to install debian (default C:\Debian)? '
from a text terminal); after it has unpacked, it will 
set up virtual mounts using an helper application; after that,
Debian/w32 base system will be functional, and the package installation
will be done from dselect (or whatever replacement we like more)

> (Can you tell I'm walking a thin line here ?)

(my english fails me...)


A Mennucc
 "È un mondo difficile. Che vita intensa!" (Renato Carotone)

Reply to: