Re: Where to from here?
On Thu, Mar 04, 2004 at 12:56:26AM -0800, Alex Perry wrote:
> John Goerzen wrote:
> >* We have a working gcc that is a 64-bit executable and generates
> >64-bit executables.
> As far as I can tell, compilers are not self-hosted on pure64 because
> we currently seem to be lacking some of the gcc-3.3 build depends.
> I found that "blas" (which uses g77) fails to build for me in pure64,
> yet there is a binary package in John's archive that is quite usable.
> That's what led me to try to rebuild g77 ... and notice the lack.
That's not correct. I have built the last several gcc versions from
within pure64 using that environment only. All that is missing,
build-dep wise, is gnat. Disable ada (and Pascal because it doesn't
build either) and you're set.
> In the short/medium term, I'm quite happy to have a distributed pool
> whereby there are a dozen lines that we all have in our sources.list
> so that each line corresponds to one amd64 special interest group.
> For example: bootstrap/build, xfree86, kde, scientific, ...
I'm not so pleased with that solution, since we'll wind up with people
stomping over each other's versions inevitably. We need some more
coordination I think.
> >* Latest versions of procps and dpkg-dev are b0rked.
> Yeah, what's up with that ? What can we do to help ?
Haven't had time to fix, mainly.
> My biggest concern is that pure64 is not able to rebuild itself.
> You mentioned failing to build a kernel, which is one such problem.
> A scripted search, looking only at what I've got installed right now,
Unfortunately, that list is misleading. Just because something fails
the build-dep does not mean that it is unbuildable, and that is
particularly the case with gcc. Look at debian/rules.defs -- many archs
don't build all packages there.
> lib64c6 fails build dependencies
> lib64gcc1 fails build dependencies
> lib64stdc++5 fails build dependencies
These packages aren't in amd64 anyway.
> Noteworthy items are lib64c6, gcc and libstdc++
> I suspect that, when pure64 becomes able to rebuild itself,
> the whole porting effort will become much more parallel.
I think that, excepting the kernel, it's already there.