IRC logs for #baserock for Thursday, 2018-07-19

*** jjardon has quit IRC05:30
*** jjardon_ has joined #baserock05:30
*** gtristan has joined #baserock06:27
paulsherwoodis wiki.baserock.org down for everyone or just me?08:05
noisecellpaulsherwood, it is down for me too08:08
paulsherwood:/08:17
ironfootall branchable.com08:35
*** gtristan has quit IRC08:44
*** gtristan has joined #baserock09:03
* paulsherwood has a backup09:15
paulsherwoodliw is checking, it's something to do with letsencrypt cert, should be resolved today09:15
*** jjardon_ has quit IRC10:18
*** jjardon has joined #baserock10:19
paulsherwoodi've pushed the wiki content to https://gitlab.com/baserock/wiki.baserock.org11:58
paulsherwoodmaybe we should take this as an opportunity to tidy up the content and present it via gitlab?11:59
*** toscalix has joined #baserock13:18
paulsherwoodmay be of interest here https://lists.trustable.io/pipermail/trustable-distros/2018-July/000008.html13:23
ironfootcircular dependencies within baserock? :S13:32
* paulsherwood is surprised by that... i thought we'd pulled them all apart13:32
ironfootwell, a circular dependency in definitions means you can't build it ever13:33
ironfootor used to mean13:34
ironfootI feel like this is a misinterpretation of the current definitions13:34
paulsherwoodflatmush: ^^ ?13:34
flatmushcircular dependencies mean things like13:35
noisecellironfoot, the circular dependencies have been broken using tarballs as far as we have seen13:35
flatmushcoreutils requires "join" which is part of coreutils13:35
flatmushwe get around that by adding a random "join.c" downloaded from openbsd13:35
flatmushand obviously gcc depends on gcc13:35
flatmushsome packages when updated become like this, flex depends on bison which depends on flex13:36
ironfootaha, I get what you mean now13:36
flatmushthis has to be handled with bootstrapping as is done by gcc13:36
flatmushbut in many cases currently, we just download a tarball13:36
flatmushlike for autotools13:36
ironfootI think that would make the whole bootstraping process  more complete, but more complicated at the same time13:37
ironfootand difficult to maintain13:37
flatmushit would add complexity to bootstrapping yeah, I don't see that you can really claim to be building from source otherwise though13:37
ironfootwhat about the packages that only recommend you to build from tarball releases?13:38
flatmushthat's gnu nonsense13:38
flatmushdo not trust13:38
ironfootI guess that with bst it would be easier to create more bootstrap elements13:39
flatmushdownload this tarball with random content I generated on my machine, but don't worry, it's totally fine13:39
flatmushseems to undermine the whole idea of building from git so that you know what you're getting13:39
noisecellironfoot, yes, with bst given that you are able to split runtime and buildtime dependencies the bootstrap makes more sense13:39
flatmushthe advice is to make the build less painful, I get that, but that's not what we're aiming for (or not only)13:39
paulsherwoodflatmush: well, if the tarballs only contain source?13:40
* noisecell would like to diff tarballs against their tags and check if that is true13:41
flatmushthe tarballs contain autogenerated content that I'd argue is impossible to read/audit13:42
flatmushgenerated by the host tools of a host machine which we have no control over13:42
flatmushif the tarballs contain source that's not in git, then that's an issue anyway13:42
paulsherwoodin the baserock case, the (configured) source is in git iirc13:43
paulsherwoodbut i take your point13:43
flatmushadding tarballs to git doesn't add anything13:43
flatmushwe either have to say we trust the generation of the tarballs or not13:43
flatmushI'm assuming the less things we have to trust, the better13:44
paulsherwoodoh, agreed13:44
*** toscalix has quit IRC17:24
*** gtristan has quit IRC19:14

Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!