Warren,
Thanks for the quick response.
> Python as a scripting environment isn't configured for that style of
> use, where all of the cross-platform components are shared in common
> and only the platform-specific binaries are separated.
machine:~/scratch/Python-2.4.2> ./configure --help
...
Fine tuning of the installation directories:
--bindir=DIR user executables [EPREFIX/bin]
--sbindir=DIR system admin executables [EPREFIX/sbin]
--libexecdir=DIR program executables [EPREFIX/libexec]
--datadir=DIR read-only architecture-independent data [PREFIX/share]
--sysconfdir=DIR read-only single-machine data [PREFIX/etc]
--sharedstatedir=DIR modifiable architecture-independent data [PREFIX/com]
--localstatedir=DIR modifiable single-machine data [PREFIX/var]
I have used these features to produce a Python 2.4.2 installation
that does exactly what you say it isn't configured to do. I see no
problem in this respect. Also, having AFS helps for making multi-
platform program installations thanks to the live expansion of "@sys"
to the current value of `sys` in file names by AFS. Witness
lrwxr-xr-x 1 root root 8 Apr 3 15:31 bin -> @sys/bin
The machine-dependent files are in `sys`/{bin,lib,include} for all of
the different types of systems that we have.
> Though it might be possible to pull something like that off, the
> megabytes saved per time spent wouldn't be worth the trouble.
Perhaps.
> [Aside: I don't know what prices are like in Europe, but here in the
> US disk space is running 50-70 cents a GB.
Disk prices are not the issue. Of course you can replace the current
hard drives in all of the peecees and workstations (>100 of them) with
disks twice or four times the current size, and it won't even cost a
lot to buy the hardware. There might be other costs, such as having
to install them all in the machines, probably wrecking the Windows
installs on them all in the process and having to redo them all.
Also, adding more space to a server-grade configuration is often
slightly more expensive than adding a single disk to your peecee.
And when changes are made to a program, I want to make one change,
not n different sets of the same change for n different platforms
if I can avoid it.
> Thus, a complete platform-specific binary PyMOL install uses about 2
> cents worth of disk space, and a complete build with all dependencies
> from source with object code costs about a quarter (0.25 USD).]
I believe, perhaps mistakenly, that there might be other advantages to
installing only one copy of each dependency (so that you don't have a
separate zlib for every program that uses it, for example, so that
when zlib security problems come up, you replace one copy instead of
trying to remember to rebuild every package that came with its own).
> As for the error -- either the _cmd.so binary isn't installed in the
> right place, or it is for a different platform.
Since I have only started building Pymol after completing the installation
of all the "ext" packages, and there aren't yet any files for any of the
other platforms but one, it's a case of the former.
Sure enough, _cmd.so hasn't been built at all.
"make" continues despite the following command failing in the middle of
the compilation process. I believe Makefiles should be written in such
a way that if any of the commands required to build the end product
fails, the rest of the effort stops there. I only thought I had it
going already because the "make" process went to the end without failing
on an error (that I hadn't seen, and that was somewhere in the middle
of it).
cc -shared -c */*.o ov/src/*.o -march=i486 -D__i686__ -ffast-math -Wall -ansi
-pedantic -Wno-long-long -Wmissing-prototypes -g -I{my Python 2.4.2
install}/include/python2.4 -I{my Python 2.4.2 install}/@sys/include/python2.4
-I/usr/X11R6/include -D_PYMOL_MODULE -D_PYMOL_INLINE -D_PYMOL_NO_MAIN
-D_PYMOL_NOPY -L{my Python 2.4.2 install}/@sys/lib -Wl,-rpath,{my Python
2.4.2 install}/@sys/lib -L{my Tcl/Tk/Tix 8.4.11 install}/lib -Wl,-rpath,{my
Tcl/Tk/Tix 8.4.11 install}/lib -L/usr/X11R6/lib -lpython2.4 -ltcl8.4 -ltk8.4
-lTix8.4 -lGL -lGLU -lpng -lXmu -lm
This command produces a ton of "linker input file unused because linking
not done" messages regarding .o files in layer0/. The command line makes
no sense - "cc -shared -c", "cc -c anything.o", and "-o something" being
missing anyway.
--
Atro Tossavainen (Mr.) / The Institute of Biotechnology at
Systems Analyst, Techno-Amish & / the University of Helsinki, Finland,
+358-9-19158939 UNIX Dinosaur / employs me, but my opinions are my own.
< URL : http : / / www . helsinki . fi / %7E atossava / > NO FILE ATTACHMENTS
