Hi, > Binpkg performance is acceptable albeit not blazing fast for machines > with 500-800 packages (usually server) while for desktops which easily > have 2000 packages the time to update can be hours. I take from this that the binpkg process really should be optimized somehow. Without looking at the code it looks like binpkg is significantly less efficient than without.
I think Michał also hinted at this. Essentially, from the sounds of the above I suspect we're at "we're probably burning more CPU on calculating the dependency tree than merely recompiling". Which to me defeats the purpose. This may be a single-core use vs multiple cores though, so whilst it may take similar amount of time it's possible that we end up consuming less energy overall. Kind Regards, Jaco
