Re: build probs on FreeBSD

On Tue, May 29, 2001 at 04:26:14PM -0500, Karl Fogel wrote:
>...
> Right now, we have a situation in which most developers cannot build
> the head of the tree from scratch. This is Bad :-).

I've heard of some dependency issues, which were solved by manually building
"libsvn_fs" before the problem thang in question. A hassle, but it doesn't
truly prevent a build.

And I just checked in a fix for it. People can do:

$ cvs up gen-make.py
$ ./gen-make.py build.conf
$ make [with-whatever-options-or-targets]

That should do the trick.

> Can you please
> get the new system working on a branch and *then* merge it into the
> trunk, so no one else is bottlenecked on this?

Nobody should be bottlenecked any more. The system worked for me before I
ever checked it in. As I said in my first email "this works on my machine."
There isn't much more that I can do except to respond to problems if/when
they come up.

> For example, Mike
> Sussman's trying to write tests, and therefore trying to build from a
> fresh checkout, and having all sorts of problems (he's typing up a
> report on that right now).

"all sorts of problems" that could have been solved with a simple "cp". I've
checked in a file into CVS to fix this one. It is a relatively known problem
with the autoconf package. There is a longer term fix, but that comes after
we stop using our own libtool (we can simply use APR's copy of libtool).

> When the new build system works with a freshly checked out tree on a
> few different operating systems, then we'll know it's ready for prime
> time. In the meantime, there's no reason why its development should
> get in anyone else's way.

Should be ready now.

> Ben mentions that the old Makefile.am's are still in their subdirs, so

Yup. That was on purpose :-) If every went to shit, then it would be easier
to restore the old system. I'll probably nuke the .am files later this week.

>...
> is still largely in place. Temporary reversion on the trunk should be
> easy.

I don't believe it is needed, and is probably more hassle than it is worth.

> Don't get me wrong -- it's great that you took this on, and (though
> there are advantages/disadvantages to each way) I'm also +1 on going
> with the single-Makefile system. But there's no technical reason why
> everyone needs to feel the pain. It can just be you. :-)

But that is the point. I'm not feeling any pain. I didn't "turn it on" until
my system was running. So I can't go any further with it.

Nobody is going to check out a branch and try it. Leave it in the HEAD so
that people will use it and we find any issues. I suspect that more people
can build than not.

I also have some free time now (not much last weekend), so I'll be working
on it some more tonite. Specifically, clean up some of the "check" stuff,
deal with the fs-helpers thing, and wrap up the doc targets.

Cheers,
-g

-- 
Greg Stein, http://www.lyra.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org