| teward | i think i discovered an evil in the tar package in Bionic, in that later gitbuildpackage pristine-tar generated items with Tar 1.30+ fail to work with < 1.30 | 00:51 |
|---|---|---|
| teward | https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=897653 is the relevant report in Debian, just hit this for something that the STudio team is working on | 00:51 |
| ubottu | Debian bug 897653 in tar "tar 1.30 breaks pristine-tar" [Grave,Fixed] | 00:51 |
| Unit193 | I have 1.30+dfsg-5, it would seem. | 00:52 |
| cjwatson_ | From that bug it looks more like we need to upgrade pristine-tar a bit (>=1.44 maybe?) | 00:54 |
| cjwatson_ | Oh also tar 1.30 isn't in bionic I see | 00:55 |
| === cjwatson_ is now known as cjwatson | ||
| bryce | we ran into that tar / pristine-tar incompatibility too | 00:55 |
| bryce | our packages set up with pristine tar on eoan couldn't be built on bionic | 00:55 |
| xnox | ginggs: it's not just the resourcewarnings, which are harmless there are really test case failures with 3.7 and 3.8 too, which appear to be triggered with a rebuild. | 11:22 |
| xnox | ginggs: note that autopkgtest, with patched testsuite against previous build passed locally for me =) | 11:22 |
| xnox | ginggs: it's all forwarded upstream, opened an issue and a pull request with CI changes to show the issue. | 11:23 |
| xnox | mwhudson: pyfffffffffftw does use dask | 11:23 |
| ginggs | xnox: thanks | 11:28 |
| xnox | ginggs: mwhudson: i'm not sure, but dask might be passing under python3.8 but not under python3.7 | 11:37 |
| ginggs | Testing with python3.8 | 11:42 |
| ginggs | = 6312 passed, 395 skipped, 32 xfailed, 4 xpassed, 20 warnings in 810.41 seconds = | 11:42 |
| ginggs | Testing with python3.7 | 11:43 |
| ginggs | = 6317 passed, 390 skipped, 32 xfailed, 4 xpassed, 20 warnings in 855.78 seconds = | 11:44 |
| xnox | hm | 11:46 |
| xnox | so no failed with either?! | 11:46 |
| ginggs | looks that way, but this is dask itself | 11:47 |
| * xnox just read "Error: Quosures can only be unquoted within a quasiquotation context." in the autopkgtest of r-cran-sf and I think I need more coffee. This is worse than ruby's twiddle-wakka and python's walrus operators | 11:49 | |
| xnox | ah | 11:49 |
| * xnox was in pyfftw contex | 11:49 | |
| * xnox was in pyfftw context | 11:49 | |
| xnox | nah, pyfftw really fails with python3.8 | 11:51 |
| ginggs | xnox: did you see pyfftw armhf fails in other exciting ways? probably #917095 | 11:52 |
| xnox | ginggs: lovely | 11:59 |
| === ricab is now known as ricab|lunch | ||
| === dannf` is now known as dannf | ||
| === ricab|lunch is now known as ricab | ||
| mdeslaur | tseliot_: hi! I'm looking at bug 1852315...it seems those three CVEs are for the vGPU software...is that actually packaged in our driver packages? is it the same thing? | 14:48 |
| ubottu | bug 1852315 in nvidia-graphics-drivers-435 (Ubuntu Eoan) "CVE‑2019‑5696 5697 5698" [Undecided,New] https://launchpad.net/bugs/1852315 | 14:48 |
| teward | can anyone assist the Studio team with an oddness? They've got a lintian ERROR which triggers a fail on an unsafe symlink which they've tried to adjust/patch the change in here in their repository, but dpkg-source barfs. They are trying to change a symlink's target by editing a symlink, not sure if there's a workaround for that. As such, however, I can't sign/build/upload the code that Eickmeyer wants uploaded (raysession). | 16:46 |
| teward | if anyone has a few moments to poke and see if they can eve do their symlink patch fix, that'd be great. https://code.launchpad.net/raysession is where the code is, everything's in master (and you can get the pristine-tar for 0.8.1 but you'll need tar >= 1.30 to make it work) | 16:47 |
| teward | cjwatson: if we backport 1.30 into Bionic it'd work, I think. I did this backport from Eoan in a PPA and made it work for pristine-tar in Bionic. So the issue is the tar package :P (wrt the tar version and pristine-tar) | 16:48 |
| teward | i'm still working on the whole Backports thing with Laney but that's a task for 2020 at this point - helping Studio during dev cycle's taking a bunch of my cycles. Also working with a few clients as a consultant on things :P | 16:49 |
| cjwatson | raysession could fix up the symlink in debian/rules build and restore it to the as-in-orig-tarball state in debian/rules clean. That's the usual workaround for this sort of thing. | 16:50 |
| teward | Eickmeyer: ^ | 16:50 |
| cjwatson | Or even just put the right thing in place after the upstream 'make install' or equivalent runs, depending on whether it's used during build. | 16:51 |
| cjwatson | In fact in that case you could just use dh_link to do it. | 16:52 |
| cjwatson | Haven't looked at this package in any detail, but that's the general range of options available. | 16:53 |
| Eickmeyer | cjwatson, teward: Thanks. Honestly, not sure how to do that. Do I make a debian/raysession.links and use dh_install_override to remove the bad link? | 16:56 |
| cjwatson | dh_link removes anything that's already there all by itself | 16:56 |
| Eickmeyer | cjwatson: Ok, excellent. I'll go that route then. Thanks! | 16:56 |
| cjwatson | np | 16:57 |
| teward | thanks cjwatson | 16:57 |
| teward | as for the pristine-tar issue the issue DOES trace back to `tar` unless we need to patch gbp / pristine-tar, but simply backporting tar seemed to fix my issues. | 16:57 |
Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!