On Aug 19, 2004, at 1:16 PM, Martin Costabel wrote:
Alexander K. Hansen wrote:
I don't recall having seen this specific issue discussed before here:
It was mentioned a couple of times, as an example why the "identical .deb" policy, if taken too literally, can lead to absurd results (like most policies).
As part of its build process the atlas package autodetects the number of CPUs, and applies a build flag appropriately. Unfortunately, the result is that the .deb files aren't identical on different machines:
Single processor G4:
-rw-r--r-- 1 root admin 5754496 19 Aug 11:02 atlas_3.6.0-1_darwin-powerpc.deb
Dual processor G4:
-rw-r--r-- 1 root admin 26614516 18 Aug 22:17 scilab_3.0-4_darwin-powerpc.deb
Given the nature of the package, it would probably be a good thing to have separate single- and dual-cpu variants (for that matter a G5-optimized one would be good, too). Presumably this would require dependencies new virtual packages for the type and number of CPUs.
You would need different variants for G3, G4, G5. Where is the problem that is being solved by such a diarrhea of packages? The very idea of atlas is that it adapts itself to the machine on which it is supposed to run, so it wants to be different on every machine.
-- Martin
Actually, I was thinking about the principle of the thing: If there's a package that produces different debs on different machines, then the policy isn't ironclad. In this case, what's the problem with adding G5 optimization or other tweaks that people on the lists are asking about?
--
Alexander K. Hansen
Associate Research Scientist, Columbia University
visiting MIT Plasma Science and Fusion Center
Levitated Dipole Experiment
175 Albany Street, NW17-219
Cambridge, MA 02139-4213
------------------------------------------------------- SF.Net email is sponsored by Shop4tech.com-Lowest price on Blank Media 100pk Sonic DVD-R 4x for only $29 -100pk Sonic DVD+R for only $33 Save 50% off Retail on Ink & Toner - Free Shipping and Free Gift. http://www.shop4tech.com/z/Inkjet_Cartridges/9_108_r285 _______________________________________________ Fink-devel mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/fink-devel
