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

[Michael Raskin]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Peter Simons wrote:
| Michael Raskin writes:
|
|  > Except that in insane Makefile replacing -Wall that is only used as a
|  > parameter to gcc meaning warning level with -Wno-all will change
|  > resulting binary.
|
| I beg your pardon?
Well, it will be a bit tricky. First, you need to set -Werror. Second,
you need to ignore compilation errors. Third, you need to specify a
wildcard expression as compilation parameters (possible in inner Make
syntax).

Now, whether I'd want to use software whose author can do all that with
Makefile (so how does he abuse C and how much will the software abuse my
brain?) is another question.

|  > | Nix doesn't know that.
|  >
|  > And no program can know such things absolutely reliably..
|
| There is no need to be defensive. Instead of being in denial about a
There is - if a really core concept is at stake..

| missing use-case, it's much more productive to find ways to improve the
| software. For example, it could be possible for a Nix expression to
| declare certain parameters as "pure", meaning that they don't need to be
| incorporated into the hash. An expression "foo" that supports make's -j
| flag, for example, might want to receive an argument "cpu_count". Still,
| instance of (foo 4) and (foo 8) are the same thing. It's not like
I think that core Nix feature could be something that just lets store
owner to install "expression A to the path that would fit expression B".
~ Then maybe a wrapper in NixPkgs could use it for setting job parallelism.

Note that Nix must not decide what is safe; it should be left to
administrator, although I am not sure if it can be allowed to users in a
sane way.

| something like that couldn't be supported in Nix. The only thing that
| might stand in the way are egos, because you can't fix problems that you
| can't admit to exist.

The question is not whether a problem exists. The question is whether
all natural solutions undermine the original design goals.

First wanted property of Nix: if a packager doesn't care much, he either
gets a correct package with all dependencies listed right etc or an
error message.

Nix security property to have: whoever gives nix-daemon an expression,
with can just build it as a low-permission user and put in store. If an
expression is malicious, we can still build it - a user would get the
same result in ~. But if someone else needs an output for some
expression and it turns out to be in the same location, we can
relatively safety assume that someone else also wahnted to build the
same exrpressions.

|  > | Which makes me wonder: is it possible to get SVN access? It feels
|  > | like posting code and/or patches to the mailing list is not the
|  > | best way to participate.
|  >
|  > Write to Eelco Dolstra..
|
| I did, and I was asking him this very question again because I received
| no response. Please don't get me wrong, I don't mean to appear
| impatient.
Maybe privmsg him once he is on #nixos when he is logged in from the
University and is actively discussing something. That way your message
will get through when he is in trusted subnet and is not completely
absorbed with some other task.
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.8 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iQEcBAEBAgAGBQJIWE7NAAoJEE6tnN0aWvw3qcQH+gOTFiVxbNFHHDBIcyi4Fw2v
AGYLm7+nNvFrWvilbgTGt8sIBxcN0aryhp0hW+6b5Ms7VZ/+sXJyuFaXLBYcWcW3
07y5VleGEgRZPKgQ+INRWv0PAcxfxs3eZ2/LYAcuQp2ZRVAeIgWqLIo2kIOTs/E/
M1MVP22cCIkOcjUzF3LY6vJ4KyW4UST6vb8v5XwuGSjcEMOIHdVNbb5kUFUQoqZK
ZrCYzEpgyR/Zh/diq2/7bWIACXPhEmCfQSHHfm9i6N37hYcUGjIt8qkTghRbWlw+
0jQ+P85XOm/NlHJkpNUr8B/PK5bm/oEF9AFNwwIvkOlgMkNj/XAx/LQaWdFi0M0=
=E2cA
-----END PGP SIGNATURE-----