Re: ssh success
On Mon, Oct 23, 2000 at 12:26:43PM +0200, Robert Bihlmeyer wrote:
> Marcus Brinkmann <Marcus.Brinkmann@ruhr-uni-bochum.de> writes:
> > I compiled openssh and friends a while ago, but never cleaned up my patches
> > to make them suitable for upstream. If you want to do that, that would be
> > great.
> openssl is polished now, the patch to the current Debian source
> package is at
That's short :)
> Marcus, is it correct to check if DEB_HOST_ARCH is defined, and fall
> back to "dpkg --print-architecture" if not? Or should I use
> dpkg-architecture -qDEB_HOST_ARCH in all cases? Or just when
> DEB_HOST_ARCH is undefined?
By now, using dpkg-architecture unconditionally is fine... it exists long
enough to make this feasible. Please do so. The correct thing to do is:
DEB_HOST_ARCH = $(shell dpkg-architecture -qDEB_HOST_ARCH)
a = $(DEB_HOST_ARCH)
or, better, replace all currences of $(a) with $(DEB_HOST_ARCH).
This is necessary to get make variables vs environment variables to
> Once I'm sure on this issue, I'll send the patch to the maintainer.
Thanks. But please also add a rule to Configure for vanilla Hurd, without
Debian. We want to make sure it works without running the Debian scripts,
too. For this you have to make sure that the architecture string Configure
determines automatically makes sense and copy the rule for debian to this
> A document that explained all the details of the architecture stuff
> (especially in connection with the hurd and/or crosscompiling) would
> be a Very Good Thing. It'd also make a fine target to point
> maintainers at in the inevitable "please don't use dpkg
> --print-architecture" bugreports.
The --print-architecture stuff is dead, although prominently used. It's all
in the packaging manual and dpkg-architecture(1). In case of doubt, just ask
me or on the list.
> I hope we can lay dpkg 1.6.999 to rest, soon...
You can go back to dpkg 184.108.40.206 or so, which is still at ftp.debian.org....
> > Well, this needs to be fixed with realloc() and a loop.
> Indeed, I just wanted to make sure that it would not loop on errors
> other than ENAMETOOLONG (the name of which I only found out later).
> I have now done the loop thingy in two occasions, and use fixed
> lengths on two others, where only a limited number of characters is
> used anyway (keyword: "%.400s").
Well, okay :)
> > There are gnome-libs packages available, please try if this works.
> They are not installable, due to:
> libgtk1.2: Depends: gconv-modules but it is not installable
Eh, just force it and try to build the gnome stuff. We really don't want to