IRC channel logs

<zimoun>thanks for the Python 2 removals push :-)

<zimoun>civodul: your stagiaire is making progress about repro2docker, right? :-)

<civodul>zimoun: yup, as you noticed!

<zimoun>yeah, really cool!

<civodul>we had that obnoxious reply about the GPL'd install script: https://github.com/jupyterhub/repo2docker/pull/1048#pullrequestreview-698655561

<civodul>it's ridiculous, but we can do like they propose

<zimoun>:/

<civodul> https://github.com/jupyterhub/repo2docker/pull/1048#issuecomment-874073887 <- Nix vs. Guix

<dstolfa>civodul: i will be trying guix-jupyter on a foreign distro soon, so i'll let you know if i run into the same problem i did on my local guix system

<civodul>dstolfa: ah yes, i need to take a look

<zimoun>civodul: hehe! Nix vs Guix, again :-) A kind of FAQ could be worth, maybe.

<civodul>yes, we keep saying that :-)

<civodul>someday we'll actually do it

<zimoun>yeah, perhaps someday ;-)

<zimoun>Julia test suite is so long…

<civodul>yeah

<civodul>1.6.1 is coming?

<zimoun>pff! I am checking the recent patches by Jean-Baptiste

<zimoun>arf! A person in my lab had the good idea to turn off the machine I use… Paf! restart all the process.

<zimoun>I could be nice to have an option something like ’-K’ to avoid to redo all from the begin. It would be also helpful when developping

<civodul>sure but...

<zimoun>but… <missing-explanations> ;-)

<civodul>in short, it's not possible :-)

<civodul>there's no way you can suspend and resume a build

<civodul>and it would be a serious security risk if you did so

<dstolfa>civodul: still getting the timeout, even on an ubuntu 20.04 server... i'm either setting something up wrong, or something is broken

<zimoun>civodul: I am half-convinced… The security is not an issue because the user know what it does and yes they can shoot in their foot. However, it could be complicated to start from a previous build and only build the difference.

<civodul>zimoun: that would need a longer explanation, but the whole security model of the shared store would collapse if this were allowed

<civodul>that amounts to giving right access to the store to all users

<civodul>s/right/write/, even :-)

<civodul>dstolfa: i can reproduce the problem

<civodul>the weird thing is that the Guix kernel does start, but everything looks as if it did not receive any message from Notebook

<civodul>i'll investigate...

<dstolfa>yeah, that's the same thing i'm having here

<dstolfa>civodul: thanks, much appreciated!

<civodul>in the meantime, you could try to time-machine to an older commit to try to see when breakage happened

<dstolfa>civodul: i can do that

<dstolfa>civodul: i can confirm that f1e0e5fde34ff0e882247498ecf78106c73284d6 is good

<dstolfa>so the regression must have happened in the last month

<zimoun>civodul: why? if a seed is added to tweak the output hash, I do not see the issue. Maybe I miss something.

<dstolfa>civodul: nevermind, i launched the python kernel by accident!

<dstolfa>it's also bad

<dstolfa>need to go back further

<zimoun>Arf! Rebased the v2 version and now test suite fails… after <too-long-time>. Anyway!

<dstolfa>68c9e0a56e008f19427bd213cf5b24bdd8fe5922 is also bad

<dstolfa>this seems to be a very old regression

<dstolfa>hmm, i wonder if i'm doing the time-machine thing right :)

<dstolfa>maybe not

<dstolfa>right, i was accidentally running it on the current commit, was using time-machine wrong

*dstolfa tries again

<dstolfa>civodul: 68c9e0a56e008f19427bd213cf5b24bdd8fe5922 is a good commit

<dstolfa>if you want to bisect, that's the place to start probably

<dstolfa>i'll try to narrow it down a bit more

<civodul>dstolfa: awesome, thanks! you can send that to the issue

<drakonis>hahaha welp, i guess nix vs guix will begin to come up a lot more often now

<dstolfa>civodul: will do!

IRC channel logs

2021-07-05.log