| 7 Jun 2021 |
 maralorn | Just for the testPhase. | 13:24:07 |
| maralorn | The only practical harm that can come from this is, that some tests pass without the author knowing that they are making assumptions. | 13:24:50 |
 cdepillabout | cabal and stack also run tests in slightly different environments. I've seen this cause weird problems with test, specifically doctest. | 13:29:56 |
| cdepillabout | I guess it could be nice to have our testing environment be more like the environments used by cabal or stack. | 13:30:25 |
| cdepillabout | * cabal and stack also run tests in slightly different environments. I've seen this cause weird problems with tests, specifically doctest. | 13:31:07 |
 sterni (he/him) | we could also contemplate using installCheckPhase | 13:31:14 |
| sterni (he/him) | this should at least mean the binary is in path all the time | 13:31:21 |
 siraben | why is tensorflow building now oh god | 13:33:26 |
 joe (he/him) | In reply to @siraben:matrix.org
why is tensorflow building now oh god RIP computer | 13:34:29 |
| siraben | Mem: 62Gi 48Gi 11Gi 8.0Mi 3.4Gi 13Gi | 13:34:47 |
| siraben | and 32 cores are maxing out... | 13:34:50 |
| siraben | * total used free shared buff/cache available
Mem: 62Gi 48Gi 11Gi 8.0Mi 3.4Gi 13Gi
| 13:35:07 |
| siraben | * total used free shared buff/cache available
Mem: 62Gi 48Gi 11Gi 8.0Mi 3.4Gi 13Gi
| 13:35:38 |
| maralorn | There seem to be issues with ghc 8.10.5 (e.g. https://gitlab.haskell.org/ghc/ghc/-/issues/19950) in #ghc:libera.chat people are talking about a 8.10.6 release to fix that … | 13:35:45 |
| sterni (he/him) | if it is related to xcode we may dodge that | 13:37:45 |
| maralorn | I am not sure if our darwin issues are the same. | 13:37:54 |
| maralorn | Just wanted to mention that it might be an upstream issue. | 13:38:17 |
| sterni (he/him) | yeah ty | 13:38:24 |
| sterni (he/him) | we'll see as soon as GHC compiles | 13:38:32 |
| * sterni (he/him) watches the build log some more | 13:38:40 |
| maralorn | In reply to @cdepillabout:matrix.org
I guess it could be nice to have our testing environment be more like the environments used by cabal or stack. Alltogether I think being a bit more lenient than the setups in which developers do tests wouldn‘t be problematic as long as we discover when the program is actually broken. | 13:47:37 |
| sterni (he/him) | In reply to @sternenseemann:systemli.org
John Ericson: I wonder whether this actually did or did not happen in the meantime since we still use those flags: https://github.com/NixOS/nixpkgs/issues/46814#issuecomment-424792721 should still work I think https://gitlab.haskell.org/ghc/ghc/-/blob/ghc-8.10.5-release/aclocal.m4#L1837-1861 | 13:48:05 |
| maralorn | I mean most of our issues with tests stem from the fact that we are stricter. | 13:48:07 |
| maralorn | sterni (he/him): Some iconv issue seems also to be known in the #ghc:libera.chat chat. | 14:18:19 |
 John Ericson | sterni (he/him): at least, those things were messing up build != host == target or build == host != target builds | 14:41:38 |
| sterni (he/him) | hmm all native currently | 15:00:30 |
| sterni (he/him) | * hmm all native currently (the issues that is) | 15:01:33 |
| siraben | sterni (he/him): could you check if the flake github:siraben/siraben-overlay#double-pendulum-simulation is building for you on darwin? | 18:29:03 |
| siraben | I overrided the input to point to 20K commits before current master and I still have the ld error | 18:29:17 |
| siraben | So I'm confused if maybe master~20000 is not enough | 18:29:43 |
