IRC channel logs

<zimoun>hi!

<zimoun>in case people missed it: <https://lwn.net/Articles/874642/> It is about embed “package information” to ELF by some RPM folks. Somehow, this feature is already included in Guix package by design. :-)

<civodul>zimoun: hi! hadn't read the article; note that we miss the CPE name, for instance

<zimoun>civodul, yes. But this information could be added as ’properties’ at package time, isn’t it?

<civodul>zimoun: yes, but for that the ideal way is to map the store item back to its package

<civodul>which we can do with provenance data in manifests

<zimoun>why is the store item required?

<zimoun>therefore, these fields could be formatted by importers, for instance.

<civodul>the use case described in the article is when you have nothing but the binary, or almost

<zimoun>yeah, but usually, the binary alone is useless. Other said, the issue that article is trying to tackle is because old-distro paradigm, IMHO. Somehow, if Guix adds a tiny field to ELF, something similar to .note.gnu.build-id, only providing the Guix commit, then done! Modulo channels information.

<civodul>yes, but i don't think we'd want to store provenance metadata everywhere

<civodul>notably because it's silent yet would involve rebuilds at each commit :-)

<zimoun>yes, I agree. :-) My comment was just to say that people from old-distro paradigm are trying to add metadata to tackle issues which are not (or almost not) for “functional”-paradigm distro.

<zimoun>somehow, they are trying to hack a kind of binary transparency, from my understanding.

<civodul>yeah, agreed

<zimoun>because I have push right on guix-hpc/website, do also I have push right on guix-hpc/guix-past channel? Or is it uncorrelated?

<civodul>not sure, there are "top-level members" and members to individual projects

<civodul>i could set you up with these if you want, if that's not already the case

<civodul>for guix-past you need to be in .guix-authorizations first

<civodul>lemme know!

<zimoun>ah, I need to setup a GPG dance before… hum?!

<efraim>guix-past is gpg signed, like guix itself

<zimoun>well, that’s because I am doing some cleaning of various packages (python2- and ocaml4.07-) and currently I mainly remove them because they are broken… Since few people if not none will dig to find a working revision, these unmaintained package should be moved before they break.

<efraim>I don't have a preference for pplacer between guix and guix-past, guix-bioinformatics already pulls in guix-past

<zimoun>but in guix, it means patch for fixing, right?

<zimoun>efraim, rekado_: another example, the recent python-decorator update breaks the bionfo package ’grit’ because python2-decorator becomes broken. The question is: do we remove ’grit’ as collateral damage and move it to guix-past or guix-science? Or do we fix ’grit’ by adding an hidding python2-decorator

<zimoun>hidden

<efraim>we

<efraim>bah apostrophe next to enter

<efraim>we've added older versions for python2- packages before

<zimoun>added to guix?

<efraim>most recently for python2-cloudpickle

<zimoun>efraim, thanks.

<civodul>rekado_: i've renewed your gitlab.inria.fr account and than of other folks here

<zimoun>civodul: 2099-12-31, long life ;-)

<civodul>did i do that? i thought it was going to be 2024 at the latest

<civodul>so you received a notification?

*civodul still learning about e-bureaucracy

<zimoun>yes, 2024 is the end date for the account

<zimoun>sponsorship end date : 2099-12-31

<civodul>oh got it

<zimoun>yes, I received a notification in my INBOX :-)

<efraim>civodul: thanks!

<civodul>yw!

<zimoun>civodul: about «deleting unused links» GC phase, thanks. It helps. However, it does not really help when “guix gc -D”. Is it not possible to turn off this phase when using the option ’-D’?

<civodul>zimoun: i don't think so

<civodul>we'll need to think some more about it

<zimoun>civodul: because, it is slow on my machine that I hit ^C^C at the phase. I am doing that since ever and I have never seen an issue.

<zimoun>civodul: as I wrote elsewhere, people running “guix gc -D” knows more or less what they are doing. Therefore, if it is not possible to fully turn off the phase when -D, it appears to me really useful to have something that turn off properly instead of ^C^C; because I cannot wait 10min or more each time I do “guix gc -D” (thus using Guix is not distratcion-free ;-))

<civodul>zimoun: i know, i understand the use case, i'm just unsure how to address it

<zimoun>civodul, hehe we know we know. :-)

<zimoun>To me, bypassing the phase is the only way; especially for single element.

IRC channel logs

2021-11-16.log