--- Log opened Thu Jul 22 00:00:14 2021
00:08 -!- belcher_ is now known as belcher
01:11 -!- copumpkin [~woohoo@user/copumpkin] has quit [Quit: My MacBook has gone to sleep. ZZZzzz…]
01:37 < darosior> jonasschnelli: fwiw i had a hash mismatch too for a texlive derivation too and had to use a substitute.
01:39 < darosior> guix wizards: i can't manage to have w64 and darwin builds to $PWD/guix-build-22.0rc1/output/, only Linux ones. I tried re-building with HOSTS="x86_64-w64-mingw32 x86_64-apple-darwin18", which was successfull but i've nothing to attest for in output/ :(
02:34 < jonasschnelli> thanks darosior 
02:51 < hebasto> https://www.irccloud.com/pastebin/YierZgB1/
02:52 < hebasto> achow101: darosior: ^
02:53 < fanquake> Yea I pointed out the difference here: https://github.com/bitcoin-core/guix.sigs/pull/39
02:54 < hebasto> it's difference in hashes for `aarch64-linux-gnu` and `arm-linux-gnueabihf`
03:01 < hebasto> ^ ignore
03:07 < hebasto> darosior: re "i can't manage to have w64 and darwin builds..." -- maybe take another try in clean repo?
03:37 < darosior> Not sure how it'd help, also i'd rather find the cause than rebuilding for X hours
04:39 -!- luke-jr [~luke-jr@user/luke-jr] has quit [Quit: ZNC - http://znc.sourceforge.net]
04:40 -!- luke-jr [~luke-jr@user/luke-jr] has joined #bitcoin-core-builds
05:57 -!- cfields [~cfields@207.172.223.102] has quit [Quit: No Ping reply in 180 seconds.]
05:58 -!- cfields [~cfields@207.172.223.102] has joined #bitcoin-core-builds
06:11 < laanwj> just pushed my codesigned.SHA256SUMS, i'm seeing some differences with achow101's 
06:11 < laanwj> -1affe7e49462b6a48c46167cb182594d4cbd54ff61486e3daef75618dbc85df9  dist-archive/bitcoin-22.0rc1-codesignatures-22.0rc1.tar.gz
06:11 < laanwj> +2c592cda367ba4be4714f7eb988ce1bb69128a19772339f86940b6c4a317bb51  dist-archive/bitcoin-22.0rc1-codesignatures-2ce133bf11e0.tar.gz
06:11 < laanwj> +1affe7e49462b6a48c46167cb182594d4cbd54ff61486e3daef75618dbc85df9  dist-archive/bitcoin-22.0rc1-codesignatures-8a51da74577c.tar.gz
06:12 < laanwj> oh for some reason i'm skipping macos, apparently
06:12 < laanwj> let's remove a stale HOSTS=...
06:23 -!- gnusha [~gnusha@user/gnusha] has joined #bitcoin-core-builds
06:23 -!- Topic for #bitcoin-core-builds: Bitcoin Core build system discussions | Channel logs: https://gnusha.org/bitcoin-builds/ | To get started learning about our build system: https://github.com/bitcoin/bitcoin/tree/master/depends
06:23 -!- Topic set by orionwl [] [Tue May 25 06:15:05 2021]
06:23 [Users #bitcoin-core-builds]
06:23 [ _0x0ff  ] [ emzy    ] [ gribble      ] [ josibake  ] [ michaelfolkson] [ willcl_ark_] 
06:23 [ achow101] [ fanquake] [ harding      ] [ kalle     ] [ provoostenator] [ windsok    ] 
06:23 [ ariard  ] [ fjahr   ] [ hebasto      ] [ kanzure   ] [ real_or_random] 
06:23 [ belcher ] [ ghost43 ] [ jamesob      ] [ koolazer  ] [ robertspigler ] 
06:23 [ cfields ] [ glozow  ] [ jarolrod     ] [ laanwj    ] [ ryanofsky     ] 
06:23 [ darosior] [ gnusha  ] [ jkczyz       ] [ luke-jr   ] [ sdaftuar      ] 
06:23 [ dongcarl] [ greypw  ] [ jonasschnelli] [ MarcoFalke] [ sipa          ] 
06:23 -!- Irssi: #bitcoin-core-builds: Total of 37 nicks [0 ops, 0 halfops, 0 voices, 37 normal]
06:23 -!- Channel #bitcoin-core-builds created Sat May 22 03:59:28 2021
06:25 -!- Irssi: Join to #bitcoin-core-builds was synced in 130 secs
06:28 < laanwj> what is the difference between dist-archive/bitcoin-22.0rc1-codesignatures-22.0rc1.tar.gz and dist-archive/bitcoin-22.0rc1.tar.gz anyhow ?
06:30 < laanwj> ohh, dist-archive/bitcoin-22.0rc1-codesignatures-22.0rc1.tar.gz contains just the code signatures
06:33 < laanwj> makes sense
06:37 < laanwj> i'm curious why 'git diff' thinks all the files are binary files
07:04 < laanwj> dongcarl: the guix.sigs repo has a .gitattributes that marks them as -text -diff, any specific reason?
08:32 < achow101> The difference appears to be because I did my code signing builds as separate invocations of guix-codesign
08:33 -!- copumpkin [~woohoo@user/copumpkin] has joined #bitcoin-core-builds
08:58 < achow101> dongcarl: isn't dist-archive supposed to be skipped by guix-attest? SKIPATTEST.TAG seems to indicate that was the intended behavior
08:59 < sipa> attestations must commit to the source code they're attesting to
09:17 < achow101> dongcarl: guix-verify seems to be looking for all.SHA256SUMS but guix-attest makes codesigned.SHA256SUMS
09:17 < achow101> darosior: your attestation mismatches
09:21 < darosior> Hmm they do mismatch with yours but match with wumpus, fanquake, jonatack and hebasto
09:21 < darosior> afaict
09:21 < achow101> mine appear to match with everyone else's
09:23 < darosior> I seem to only mismatch with you on my side: https://0bin.net/paste/ZpacqUOb#rXejBw8rXsiRDJZq-njma8m4x6bRjlZ5/azaOGmL0go
09:25 < achow101> no, it's because guix-verify uses the first manifest it finds to do the comparison
09:25 < achow101> so because my name is first, it is comparing mine against everyone else's
09:25 < achow101> SIGNER doesn't do anything, although it should
09:25 < achow101> I'm working on a few fixes
09:32 < darosior> Lol ok thanks for pointing it out: https://0bin.net/paste/bylR0xej#cHVGbUrhWxLyu4pqK7FH91RgefTcDUWRWHfLTaizgue
09:33 < darosior> Will try again tomorrow, hopefully i can also manage to build darwin and w64..
09:42 -!- cfields [~cfields@207.172.223.102] has quit [Ping timeout: 265 seconds]
09:43 -!- cfields [~cfields@207.172.223.102] has joined #bitcoin-core-builds
09:43 < emzy> oh no. My build stopped at "/gnu/store/qafnqv7s048z47d4h6yfb1fc86a6fh4r-gdk-pixbuf-2.40.0.drv failed"
09:44 < emzy> " 5/21 gdk-pixbuf:slow / pixbuf-randomly-modified  FAIL     7.37 s (killed by signal 5 SIGTRAP)"
09:46 -!- luke-jr [~luke-jr@user/luke-jr] has quit [Quit: ZNC - http://znc.sourceforge.net]
09:46 -!- luke-jr [~luke-jr@user/luke-jr] has joined #bitcoin-core-builds
09:49 < emzy> not enough RAM: "failed to allocate 524288 bytes"
09:49 < emzy> do I have to start over? Or is there a way to continue?
09:50 < achow101> emzy: when you restart, it should pick up where it left off since guix will still have the packages it already built
09:53 < emzy> achow101: contrib/guix/guix-build is complaining: "ERR: Build directories for this commit already exist for the following platform
09:53 < emzy>      triples you're attempting to build, probably because of previous builds.
09:53 < emzy>      Please remove, or otherwise deal with them prior to starting another build."
09:54 < achow101> use guix-clean, it should be mentioned in the error message
09:55 < emzy> "Hint: To blow everything away, you may want to use: ./contrib/guix/guix-clean" sounds like is will start a clean run. But I will do it.
09:58 < laanwj> that cleans the build directory, IIRC the derivations are in /gnu/store so they won't be removed by that
09:58 < laanwj> as it was building derivations, it should still continue where it left off
09:58 < emzy> ok. just sounded scary.
09:59 < emzy> but also all the building of bitcoin-core will take long on my machine.
10:00 < emzy> I'm building again with more swap :)
10:00 < laanwj> having enough memory is one of the most important things for build speed
10:00 < laanwj> if it needs to swap it's going to take ages, you might want to lower parallelism instead
10:01 < emzy> It has 4GB but I had X running and only 1,6GB Swap.
10:03 < emzy> I need another dedicated machine for the builds. It is some thin client I had lying around.
10:16 < darosior> emzy: had the same issue. Maybe try reducing JOBS?
10:16 < laanwj> i really need to build my riscv pc, would be nice to do a guix build on that
11:45 < laanwj> finally ordered the components :)
11:46 < sipa> nice
11:46 < sipa> i wouldn't have too much hope that if guix building on riscv works, the results will match the others...
11:49 < MarcoFalke> [15:37] <laanwj> i'm curious why 'git diff' thinks all the files are binary files
11:50 < MarcoFalke> See https://github.com/bitcoin-core/guix.sigs/blob/main/.gitattributes 
11:50 < laanwj> MarcoFalke: but why
11:50 < sipa> 07:04:22 < laanwj> dongcarl: the guix.sigs repo has a .gitattributes that marks them as -text -diff, any specific reason?
11:50 < MarcoFalke> sipa: I should learn to read all lines
11:51 < laanwj> sipa: it will probably take some effort to get it deterministic w/ the x86_64 builds
11:51 < laanwj> but first step would be getting guix to run on risc-v
11:51 < MarcoFalke> The file was copied from gitian.sigs: https://github.com/bitcoin-core/gitian.sigs/blob/master/.gitattributes 
11:51 < sipa> start with aarch64 :)
11:52 < sipa> MarcoFalke: that one just says -text, which i guess is useful to avoid CR/LF messing (based on commit message)
11:52 < sipa> but why -diff too?
11:53 < laanwj> ooh  the cr/lf thing makes sense
11:54 < MarcoFalke> no idea why it is -diff too
11:54 < laanwj> right
12:06  * luke-jr wonders why donwloads work fine (even simple HTTP ones) yet git network activity gets stuck
12:50 -!- blkncd [sid505676@id-505676.brockwell.irccloud.com] has joined #bitcoin-core-builds
14:17 < dongcarl> I guess -diff only makes sense for *.SHA256SUMS.asc and not *.SHA256SUMS
14:18 < dongcarl> Yeah it should probably be removed for *.SHA256SUMS
14:34 < achow101> dongcarl: do you think we should rename codesigned.SHA256SUMS to all.SHA256SUMS? The only reason I kept it as codesigned.SHA256SUMS is because that's what's already in guix.sigs
14:38 < dongcarl> Yeah I do think so. I apologize for not fixing that codepath when we switched to the new hierarchy. Happy to do the renaming on the guix.sigs repo (I guess the CI won't be too happy but I'm sure someone has the rights to override that).
14:38 < dongcarl> Also, SKIPATTEST.TAG is suuuper confusing... Removing it...
14:46 -!- luke-jr [~luke-jr@user/luke-jr] has quit [Quit: ZNC - http://znc.sourceforge.net]
14:48 -!- luke-jr [~luke-jr@user/luke-jr] has joined #bitcoin-core-builds
14:51 -!- luke-jr [~luke-jr@user/luke-jr] has quit [Client Quit]
14:51 < dongcarl> achow101: Thanks for the update!
14:52 -!- luke-jr [~luke-jr@user/luke-jr] has joined #bitcoin-core-builds
14:55 < achow101> dongcarl: re the current codesigned mismatch, I suppose there isn't a way around it other than for me to remember to redo guix-codesign after tagging the detached sigs repo?
14:56 < achow101> what I had done this time was run guix-codesign prior to tagging as a sanity check to make sure the sigs attached, which is why I have extra dist-archive things with commit hashes instead of version names
15:00 < dongcarl> I'm a bit confused about your situation, let's talk offline
15:09 -!- luke-jr [~luke-jr@user/luke-jr] has quit [Quit: ZNC - http://znc.sourceforge.net]
15:10 -!- luke-jr [~luke-jr@user/luke-jr] has joined #bitcoin-core-builds
16:09 < achow101> luke-jr: I've updated the gitian descriptor so it should be caching correctly now. Also added the git stuff it fetches to the cache so hopefully you won't get stuck on that every time
16:09 < achow101> I had to remove the quotes around $STORE_FILES
16:37 < dongcarl> Talked with achow101 a bit, we’ve identified a few tooling problems in rc1 arising out of the changes after the 21.99-guixtest1 run. Here they are:
16:38 < dongcarl> 1. guix-attest produces codesigned.SHA256SUMS but guix-verify expects a different name: all.SHA256SUMS. This is fixed by uniformly using all.SHA256SUMS in #22531
16:38 < dongcarl> 2. guix-attest mistakenly “normalizes” the SHA256SUMS document twice when producing all.SHA256SUMS, resulting in a document with malformed <CR><CR><LF> line endings and triggering our `comm` sanity check in guix-verify. This is fixed by: https://github.com/bitcoin/bitcoin/pull/22531#issuecomment-885297071
16:38 < dongcarl> To avoid confusion, perhaps we should hold off on merging rc1 attestations in guix.sigs until we have an rc (2?) with these fixes included.
16:38 < gribble> https://github.com/bitcoin/bitcoin/issues/22531 | guix: Fixes to guix-{attest,verify} by achow101 · Pull Request #22531 · bitcoin/bitcoin · GitHub
17:11 -!- belcher_ [~belcher@user/belcher] has joined #bitcoin-core-builds
17:15 -!- belcher [~belcher@user/belcher] has quit [Ping timeout: 255 seconds]
17:26 < dongcarl> Okay I just ran through the entire flow including codesigning and verifying in a dirty tree (HEAD = v22.0rc1) with all the fixes. Seems like things are all fixed.
17:51 < fanquake> time to do some merging
17:51 < fanquake> time to do some backporting
17:51 < fanquake> time to do some building
22:56 -!- vasild [~vd@user/vasild] has joined #bitcoin-core-builds
23:47 < luke-jr> achow101: testing
23:48 < luke-jr> before, strangely, I enabled TCP keepalive on the VM, and saw it succeeding even while git hung apparently waiting on the socket O.o
23:48 < luke-jr> not sure what's going on with that side of things
--- Log closed Fri Jul 23 00:00:14 2021