IRC channel logs

2022-04-20.log

back to list of logs

<drakonis>civodul: that is a correct assessment about github
<drakonis>it does indeed make it easier for drive-by contributions to happen
<drakonis>it has become so prevalent that a reaasonable amount of large projects have started hosting on github because of that
<drakonis>newer linux distributions without the manpower or inertia to roll out their own infra use it
<drakonis>llvm has moved away from its own repositories and now hosts on github
<drakonis>sbcl, despite being available through sourceforge, it still uses github
<drakonis>gentoo has development done both on its own git repository and on github too
<drakonis>it has done wonders for sending PRs
<drakonis>there's also rust, which treats github as a first class source for fetching crates
<civodul>drakonis: yup, i'm well ware that avoiding Git{Hub,Lab} is a key ingredient of our "avoiding success at all cost" recipe :-)
<civodul>*aware
<civodul>it is terrible
<civodul>because as you say, it can do wonders
<drakonis>indeed it is.
<drakonis>though sourcehut is quite nice
<drakonis>the correct way to phrase it is 'avoid (success at all costs)'
<civodul>oops right
<drakonis>although haskell itself does 'avoid success at all costs' instead
<drakonis>which is a shame
<civodul>heh
<civodul>now, our problem today is lack of reviewers more than lack of contributors
<civodul>so i guess we'll have to address this one first
<civodul>well we've known this for some time already
<drakonis>a couple years now
<drakonis>yeah
<drakonis>the workflow needs a way to visualize which packages are currently being reviewed and active
<drakonis>mumi only goes half way through
<drakonis>does guix have anything to automate patch testing?
<drakonis>at least something that can be manually invoked by reviewers to check it
<drakonis>is 1.4.0 out yet?
***rekado_ is now known as rekado
<rekado>“I am currently using tensorflow==2.2 and keras==2.4.3 […] in conda […] The error that results is: […] I cannot utilise tensorflow==1.14 and keras==2.2 due to hpc limitations.”
<rekado>I wonder what “hpc limitations” these are.
<drakonis> https://lore.kernel.org/lkml/20220419135908.39606-1-cai.huoqing@linux.dev/ good news everyone
<drakonis>the nvidia scourge is repenting
<drakonis>guix might see a uptick in users when they actually publish gpu drivers :v
<civodul>drakonis: sounds like good news indeed
<civodul>let's hope there are no fine prints...
<drakonis>we'll soon find out
<drakonis>knowing nvidia, there might be some caveat
<drakonis>but it will still soften things up
<drakonis>as it would remove the need for workarounds to use packages with dependencies on nvidia libraries
<drakonis>this is the main issue
<drakonis>oops that doesn't sound right
<drakonis>i meant the overhead added by the workarounds to support nvidia's drivers and library
<drakonis>nix has some extremely nasty stuff that doesn't work quite right on standalone installs
<drakonis>i think it makes everything that depends on mesa also transitively depend on libglvnd
<drakonis>anyways, if the only caveat is requiring firmware for non critical functionality, it should be mostly acceptable
<drakonis>i'd be surprised if cuda didn't require anything besides the toolchain
<drakonis>worse yet, the hardware generation support range might work only in recent gpus