IRC channel logs

<zacts>bavier: nice, thanks :-)

<zacts>re: 'tools to help fix reproducibility issues'

<zacts>does mark_weaver ever hang out here anymore?

<Apteryx>Our lxqt-common package source redirects to https://github.com/lxde/lxqt-common-deprecated--do-not-use-anymore-/archive/0.9.1.tar.gz. Hm.

<Apteryx>I think due to a reorg of the repo.

<ng0>hey… where are bash-completions located normally on the filsystem? datarootdir/share/bash-completion/completions it is not…

<ng0>I meant: @datarootdir@/bash-completion/completions it is not…

<ng0>oh got it

<ng0>etc/bash_completions.d

<dustyweb>ACTION looks at go-names-are-dns-dependent conversation on guix-devel and thinks "gosh, what would happen if GitHub went the way of Gitorious"

<dustyweb>I'm sure plenty of people can't believe that would be possible now but many of us have lived through several high-profile forges dying

<dustyweb>heck, what if a lot of those go libraries were written on Gitorious a few years ago

<dustyweb>I guess Java has this convention too though in a sense

<thomassgn>I sort of realized last time I tried to submit a package that it should go to patches list and not dev - is that right?

<thomassgn>I'm checking and building a definition for screen-message now, and would like to send it upstream...

<dustyweb>thomassgn: that's best practice yes :)

<thomassgn>cool

<civodul>yup, see https://www.gnu.org/software/guix/manual/html_node/Submitting-Patches.html

<civodul>dustyweb: BTW, the other day i looked at the white paper you posted, and i liked it

<civodul>guile-csexp looks cool too :-)

<civodul>uh i seem to have lost the URL, was it not on gitlab.com?

<dustyweb>civodul: oh nice :)

<dustyweb>civodul: for the whitepaper or guile-csexp? :)

<dustyweb>I have the activitypub paper on gitlab but sadly the most up to date version is on github because that's where Rebooting Web of Trust is collaborating!

<dustyweb>I actually have three there

<dustyweb> https://github.com/WebOfTrustInfo/rebooting-the-web-of-trust-fall2017/blob/master/draft-documents/activitypub-decentralized-distributed/activitypub-decentralized-distributed.md <- "even more distributed activitypub"

<dustyweb> https://github.com/WebOfTrustInfo/rebooting-the-web-of-trust-fall2017/blob/master/draft-documents/lds-obcap/lds-obcap.md <- certificate chaining as a substrate for object capabilities on the web

<dustyweb>that's the one I got to work on with Mark S. Miller!

<dustyweb>and finally https://github.com/WebOfTrustInfo/rebooting-the-web-of-trust-fall2017/blob/master/draft-documents/smarm.md a much more work in progress writeup of how to use Jonathan Rees' work on security capabilities as the lambda calculus to build a language suitable for "smart signatures/contracts"

<civodul>ah yes, thanks!

<dustyweb>that last one needs the most work

<civodul>smarm.md is the one i was referring to

<dustyweb>cool

<dustyweb>yeah I need to finish it

<dustyweb>I also built a prototype in Guile that did not do the deterministic bounding

<dustyweb>but did the obcap stuff

<dustyweb>turns out

<dustyweb>guile's sandboxing module allowed for a super fast demo

<dustyweb>but it's so rough I'm not ready to publish that... it's barely any code anyway

<civodul>heheh :-)

<dustyweb>the hard part would be specifying exactly what primitives are given to the initial environment and what the "costs" of each would be... and implementing the appropriate "accountant"

<dustyweb>I also think to be deterministic about the space aspect

<dustyweb>you need deterministicly specified gc

<dustyweb>so every N steps in time, you need to do full GC

<dustyweb>maybe there is another way around it, but if you really really want to be precise

<dustyweb>I think that will be necessary

<dustyweb>that's more important if you want to precisely declare the *exact* amount of memory used

<dustyweb>well, exact memory cost ;)

<dustyweb>implementations may vary in terms of how much memory used per deterministic cost :)

<civodul>don't talk about memory usage and GC, that makes me nervous ;-)

<dustyweb>hahaha

<dustyweb>ACTION adds weak hashtables to smarm

<civodul>:-)

<civodul>but yeah, i imagine that determine what "costs" are acceptable can be hard

<dustyweb>ACTION nods

<dustyweb>and tedious :)

<dustyweb>I imagine if one were tasked with building a bounded-smarm

<dustyweb>that would be the super boring part :)

<dustyweb>and the gc would be the super frustrating part! :)

<jonjits[m]>I'm using guix package manager on top of funtoo (a gentoo distro). I am now getting a weird error when trying to install software "Cannot write: No space left on device". I have plenty of diskspace/inodes left on the device where /gnu/store is. Is it using a tmp dir potentially somewhere else that could cause this?

<mekeor>civodul: you once showed me your fix to make cups-drivers be able to use ghostscript. has this already be merged into master? this would be so nice!

<civodul>mekeor: yes it was merged a few days ago

<civodul>jonjits[m]: could it be that /tmp is out of space?

<mekeor>awesome! <3

<Apteryx>How do I insert the ^L character in Emacs?

<alezost>Apteryx: C-q C-l

<Apteryx>thanks!

<alezost>no problem :-)

<civodul>^L, the ultimate GNU hacker trick :-)

<snape>Apteryx: and there is emacs-page-break-lines (Guix package)

<snape>to display them in a nice way

<mekeor>what's ^L?

<snape>it's a page break

<mekeor>ah, i remem er

<snape>it allows one to split a big file into different pages

<snape>and then, with 'forward-page' and 'backward-page', you can navigate through them

<Apteryx>snape: sweet. Works treat with C-x [, C-x ].

<Apteryx>Hmm? ls ~/.cache/guix/pull/ -> pjmkglp4t7znuugeurpurzikxq3tnlaywmisyr27shj7apsnalwq; git fetch && git reset --hard pjmkglp4t7znuugeurpurzikxq3tnlaywmisyr27shj7apsnalwq -> fatal: ambiguous argument 'pjmkglp4t7znuugeurpurzikxq3tnlaywmisyr27shj7apsnalwq': unknown revision or path not in the working tree.

<Apteryx>I guix pulled this morning. How can this be?

<Apteryx>the git commands are ran in my ~/src/guix checkout

<lfam>Apteryx: I'm not sure how that random string is supposed to represent. Is it supposed to be a Git commit?

<lfam>I mean, what it is supposed to represent

<Apteryx>yes

<lfam>It's not even the right length

<lfam>Mine are 52 characters

<lfam>I have this one too: pjmkglp4t7znuugeurpurzikxq3tnlaywmisyr27shj7apsnalwq

<Apteryx>eh. I'm sure I used that approach successfully before.

<lfam>They are the same length as base-32 SHA256 hashes to me

<lfam>s/to me//

<lfam>You can get the effective Git commit with `guix --version`

<Apteryx>OK! I'll use this from now on, thanks!

<Apteryx>Ah, what I had done before was more like this I think: pushd ~/.cache/guix/pull/* && git log -1; popd. Sorry for the confusion!

<Apteryx>Your suggestion is much simpler though! I'll stick to it :)

<Apteryx>I've got the full list of the broken github packages... now to automate updating the hashes directly in the guix sources.

<Apteryx>Could there be a way to guix pull using a locally built guix? I find myself wanting to use 'guix pull' to safeguard a working guix not susceptible to be GC'd, but then have to compile stuff twice because I also want to keep my ~/src/guix up-to-date.

<Apteryx>Something like 'guix pull --from-checkout=.' would be nice I guess.

<Apteryx>Or maybe I could just hack something to steal the .go of the guix pull'd version into my checkout to speed up things... Eh.

IRC channel logs

2017-10-14.log