Skip to content

Use tmpdir #2

Merged
merged 2 commits into from Apr 5, 2022
Merged

Use tmpdir #2

merged 2 commits into from Apr 5, 2022

Conversation

donald
Copy link
Contributor

@donald donald commented Apr 5, 2022

Use TMPDIR so that we don't litter Linux kernels in /srcatch/local during the beequeen builds on the cluster nodes.

While we are at it, remove obsolete kernels.

Addresses #1

Copy link
Contributor Author

@donald donald left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I am not allowed to give a "Request changes" feedback on my own PR. Why?

build.sh Outdated Show resolved Hide resolved
@wwwutz
Copy link

wwwutz commented Apr 5, 2022

hmmm... still litters ... how about some trap rm_rf_kernel() EXIT magic ?

@wwwutz
Copy link

wwwutz commented Apr 5, 2022

I love it how we hack 'quick&dirty-fix' down to almost perfect software. btw, I'm missing a man page... 8-)))

@donald
Copy link
Contributor Author

donald commented Apr 5, 2022

still litters

I though you mxqsub with --tmpdir ?

@wwwutz
Copy link

wwwutz commented Apr 5, 2022

yes... but that's a mxq feature 8-) my bad feelings come with 'oh, let's build some itty bitty tiny kernel module' and leave xx GB behind. somewhere... you would notice and complain if that'll be $CWD or $HOME ... anyway, not that of a problem. but you stumble upon a lot of extracted kernel-trees in /scratch/local{,2}, /dev/shm etc. nowadays.
make clean maybe ? 8-)

@donald
Copy link
Contributor Author

donald commented Apr 5, 2022

  • Cluster build with --tmpdir cleans up automatically now that TMPDIR is used.
  • btw: Don't you have the same problem with the nvidia_linux-* bee packages? Don't these extract to /scratch/local as well?
  • I usually build on my workstation and litter my /srcatch/local and clean it up when it gets full. Until then I appreciate the cache, because I might rerun the scripts. That's why there is a -d test.
  • That there are other users who don't clean up their waste in /scratch/local on public machines is a general problem, not specific to this package.
  • Auto-cleanup is never reliable.
  • I would a "make clean" expect to clean up the build directory I am in not a (shared) cache of extracted kernels in another path. An we don't even have a make file yet.

Is this PR an improvement? Take it as is and maybe clean up /scratch/local/linux-*.x86_64 on all systems once?

@wwwutz
Copy link

wwwutz commented Apr 5, 2022

ok ok ok ... sorry... I went over the top. $TMPDIR fixes my issue 8-) Everything else was nosepicking 8-)

This PR isn't an improvement... it's the genius perfect most high level godlike PR I've ever seen for what I have asked for. I love it. I would have merged immediately ( but as you know I don't like merging other ppls PRs).

sorry for the noise, I was distracted.

@donald
Copy link
Contributor Author

donald commented Apr 5, 2022

genius perfect most high level godlike PR

Ah, the chit-chat finally paid off.

@donald donald merged commit e4c41a4 into main Apr 5, 2022
Sign in to join this conversation on GitHub.
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

3 participants