--- Log opened Wed Jul 03 00:01:00 2019 07:37 -!- hebasto [~hebasto@95.164.65.194] has joined #bitcoin-builds 08:35 < nothingmuch> dongcarl: i would add a note, e.g. "Follow the Guix installation guide, *making sure to allocate enough disk space to the store*", since several people ran into problems 08:36 < nothingmuch> unfortunately i'm not sure what "enough" is at the moment, currently rebuilding after guix gc 08:46 < hebasto> nothingmuch: +1 08:47 < dongcarl> nothingmuch: Great. Lmk how large the store end up being 08:58 < nothingmuch> dongcarl: kind of a nitpick but instead of --nproc=4 and a "change accordingly" comment, why not $(nproc)? 08:59 < dongcarl> nothingmuch: Yeah I wanted to do `$(nproc)`, but realized that not everyone has enough RAM 08:59 < nothingmuch> ah good point 09:00 < nothingmuch> i wonder if it makes sense to mention the --profile argument, or if that just complicates things (i.e. if someone is using guix for anything else, they probably already have their considerations and would know about it) 09:00 < dongcarl> `guix` already does `-j$(nproc)` for each job by default... multiplying that by `$(nproc)` jobs at once... will be tough 09:01 < nothingmuch> wait for the guix pull? shouldn't that override instead of multiply? 09:01 < dongcarl> nothingmuch: There's a difference between `--cores` and `--max-jobs` I think? 09:02 < nothingmuch> ah, i was wondering why both exists 09:02 < dongcarl> As in `guix` jobs = jobs at a package level = how many packages to build at once 09:03 < dongcarl> Yeah I think those who use `guix` already will know about `--profile`... 09:04 < dongcarl> Might be a good tips and tricks? 09:04 < dongcarl> I'll think about it 09:09 -!- MarcoFalke [~none@198.12.116.246] has quit [Read error: Connection reset by peer] 10:30 -!- hebasto [~hebasto@95.164.65.194] has quit [Remote host closed the connection] 13:11 < dongcarl> good news everyone... It seems we can actually remove all the faketime wrapper things... 13:28 < nothingmuch> cool! is that the new changes in your guix branch? i didn't look at that yet 13:28 < dongcarl> nothingmuch: Nah I'm messing around with the `build.sh` and seeing what I can get away with without breaking reproducibility... 13:28 < dongcarl> It's going to be much tighter 13:28 < dongcarl> will push up at some point tho 13:30 < nothingmuch> so was that just carried over from gitian before other sources of nondeterminism were fixed? 13:30 < dongcarl> nothingmuch: Yup 13:31 < dongcarl> Just my strategy of `git bisecting` what was causing nondeterminism haha 13:35 < nothingmuch> cool 13:36 < nothingmuch> fwiw i suspect the gzip wrapper can also be removed, since make dist invokes it separately from tar 13:37 < dongcarl> nothingmuch: thanks for the heads up, will look into it 13:42 < nothingmuch> uh, so that build i thought i had started before i went off for a few hrs - i forgot to update to the latest copy of your branch 13:42 < nothingmuch> if i don't screw it up again i should be able to report storage reqs by tomorrow 13:42 < dongcarl> nothingmuch: roger that, thanks! 13:48 < nothingmuch> i feel esp. silly since i actually read all the diffs yesterday, i just didn't do a checkout 13:49 < dongcarl> nothingmuch: :-) Heh I'm glad you read them! 13:51 < nothingmuch> well, it's really important work, plus i have many personal reasons to care about it, so seems like good way to stop distracting myself with only superficial familiarity with bitcoin stuff 13:51 < dongcarl> nothingmuch: :-) Fantastic, wonderful to have you onboard! 13:52 < nothingmuch> i realized i've been studying this crap for like 3 years now, some of it almost full time, and there's really no excuse for waffling so much... 1st year was wasted figuring out why all the alternatives were bunk, but the rest of it is just getting distracted 13:52 < nothingmuch> so much interesting stuff going on, but it's time to switch from a BFS phase to a DFS one 13:54 < dongcarl> nothingmuch: What were you focused on those 3 years, if it's not rude to ask? 13:55 < nothingmuch> happy to answer, actually 13:56 < nothingmuch> although, maybe switch to a venue where this doesn't spam everyone else? 13:57 < dongcarl> sure! 21:18 < nothingmuch> so starting with a non clean guix install (pulled a few guix updates, installed a handful of packages to main profile), /gnu/store grew from 2.7G to 5.4G 21:22 < nothingmuch> hmm, don't think guix size is able to report on packages from a custom include path as it stands 21:30 < nothingmuch> i have emacs & gnunet installed, both of which have non trivial dep sizes, curiously guix size on all packages from my profile is 1.5G total 23:09 < nothingmuch> hmm, strangely egrep -o '/gnu/store/[^/"]+' $GUIX_ENVIRONMENT/manifest | sort -u | xargs du -shc only reports 1.4G in total 23:10 < nothingmuch> i guess extra size is for native inputs of these, e.g. the gcc bootstrap chain 23:15 < nothingmuch> and looking at guix gc --list-roots i think that 2.7G was an overestimate - guix gc --list-roots showed me some older stuff, gc'ing that just cleaned another 1.4G which is more in line with guix size estimates --- Log closed Thu Jul 04 00:00:00 2019