Hooray for Paul, who's keeping alive Swarm.
Il 27/04/14 07:28, Paul Johnson ha scritto:
> Hooray.
>
> I thought this was going to be like the old days when I had to write
> and beg for help. Not so fast. I
think I've got it.
>
> Today I've been building and testing Swarm-2.4.1 on various Ubuntu. I
> believe all problems are solved, I've written out all details here. If
> these fixes don't work for you, let me know and I'll try to replicate.
>
> I wrote it out here
>
>
http://pj.freefaculty.org/blog/?p=263>
> In the end, the Swarm apps still can run. There are shared library and
> compiler flag woes. But alls well that ends well.
>
> One must replace the blt packages provided by Ubuntu with the ones I
> offer. This is necessary because Swarm is trained to tcl/tk 8.5 and we
> need BLT to keep its attention in that version. It must not wander
> over to the 8.6 side of things. If you ignore this advice, the error
> you get is one that we have seen since 1997: TKExtra instance: no such
> key
>
> Another wrinkle is the Debian/Ubuntu multilib directory structure.
> Now, They use /usr/lib/i386-linux-gnu and /usr/lib/amd86_64-linux-gnu.
> The Swarm configure statement in 2.4.1 was set to look
for dynamic
> libraries in /usr/lib, /usr/local/lib, /usr/lib32 and /usr/lib64. I
> took the easy road by just editing the Swarm configure script and
> putting in the path. One of my long-standing embarrassments is that I
> just don't understand GNU automake/autoconf setups and I don't know
> how to make those changes in the underlying specification. But I'm
> pretty sure somebody who does understand that would understand what
> I'm saying and know what to do.
>
> The other wrinkle is the gcc linker. I can't understand why this
> crops up just now, but the compiler is
stricter. If one uses a math
> function like pow or sin, it was not previously necessary to
> explicitly link in the math library. gcc would just find it. They
> have eliminated that convenience, and "-lm" must be added as a flag.
>
> pj