[Nix-dev] Re: make -j<n> howto?

[Isaac Dupree]

Arie Middelkoop wrote:
> Eelco Dolstra wrote:
>> (Undoubtedly Gentoo people regularly build packages with specific -j 
>> flags, so they probably know how much breakage it causes...)
> 
> I'm a Gentoo user who updates frequently from the official portage tree, 
> using -j 3. I never noticed make problems due to race conditions, 
> although I'm not sure if I would be able to recognize one.

There are a lot of breakages you won't notice when using Gentoo, because 
specific packages filter certain common flags that are known to cause 
breakage in that package (for example, -O3, make -j ...) so they 
silently work without using your exact flags.  (Which is very convenient 
as a gentoo user... which I was a few years ago)

On the other hand, (just having read dolstra-thesis.pdf), I think it 
would be interesting to see how much breakage it tends to cause.  Sure, 
it's nearly impossible in general to prevent builds from being 
nondeterministic -- but can't we observe whether particular packages 
tend to compile the same way every time (to the exact same NAR/hash), 
which lets us do all kinds of interesting comparisons? Probably, making 
GCC's __DATE__ etc. not return the actual date, would accomplish that 
for most packages?  Has this been investigated at all?  (Does current 
Nix use the intensional or extensional model?)

-Isaac