Page 1 of 1

native_client (NaCl) and Debian packaging

Posted: Sun Dec 28, 2014 5:31 am UTC
by terminal

Hi,

I attempted to build the game (3a492ec56f4ffadbdc48d00da4e5d3bb44e5a61a) in a clean network-disconnected chroot and I noticed that it has a semi-hidden dependency on Google's Native Client library ("nacl-sdk"), which is currently not packaged in Debian.

Next I tried to build nacl-sdk from source, but that failed too as its build script attempted to pull in a bunch of tarballs (including newlib, clang, and llvm) over plain unauthenticated http.

After a bit of googling, I found out that nacl-sdk is part of Chromium (the "native_client_sdk" subdirectory of the source). But Chromium is packaged in Debian! Wishful thinking dictated that whoever packaged Chromium must have also adapted the nacl build script to also use the Debian sources (instead of downloading random unsigned tarballs). Wishful thinking was wrong, of course, since the Chromium nacl-sdk build fails. I'll attempt a full build of Chromium (which will likely take a few hours) and see if nacl somehow gets built correctly as a byproduct (oh look, more wishful thinking).

So anyway, what are the future plans for NaCl? Are there plans to split Debian's Chromium package to build libnacl{,-dev} out of it?

Regards,
terminal


Re: native_client (NaCl) and Debian packaging

Posted: Sun Dec 28, 2014 7:41 pm UTC
by kangz

Yeah the NaCl SDK is a very obscure thing that is a pain to get and compile. Only one of our developers went through the trouble of getting the different things we need and put them in the tarball, it would be nice if there was the steps taken to build these binaries somewhere on the wiki. So I don't think we can provide a good solution to this, and we don't have enough leverage to splitting nacl from the chromium package.

What we should do though is sign the tarballs because you're right that unsigned tarballs over http is a bad thing.


Re: native_client (NaCl) and Debian packaging

Posted: Sun Dec 28, 2014 10:23 pm UTC
by Viech

Also see the relevant issue.