| 4 Nov 2024 |
 lxsameer | it is failing on quickcheck now which I'm investigating | 16:30:41 |
| lxsameer | ok it seems that dontCheck function will not prevent the dependencies to run their tests | 16:40:42 |
 maralorn | In reply to @lxsameer:matrix.org
ok it seems that dontCheck function will not prevent the dependencies to run their tests Yes, definitely. | 16:41:13 |
| maralorn | In reply to @maralorn:maralorn.de
Is primitive in your lock file? Sorry I meant the freeze file. | 16:41:49 |
| lxsameer | do you have any suggestion on how to approach this problem with dependencies and tests? | 16:44:16 |
| maralorn | In reply to @lxsameer:matrix.org
do you have any suggestion on how to approach this problem with dependencies and tests? I don't get exactly what's the problem, rn. | 17:21:34 |
| maralorn | I thought you use your override to disable test on all deps that you override. | 17:22:15 |
| maralorn | Are you doing that an is it still failing? | 17:22:37 |
 pwmosquito | not sure if anyone has any opinion on it but the active-repositories: none Cabal issue is being discussed here:
https://github.com/haskell/cabal/issues/10504
for reference i've used this for years to essentially tell Cabal "build with the packages i provide" or in other words separate building (with Cabal cause it's incremental) from package management (cause we use Nix for that)
| 17:22:47 |
| lxsameer | I'm using hasql it is indirectly depends an selective and selective uses a quickcheck version for testing which is not compatible with the one in the freeze file | 17:22:57 |
| lxsameer | In reply to @maralorn:maralorn.de
Are you doing that an is it still failing? yeah since dontCheck only applies to hasql and not it's indirect dependencies | 17:24:25 |
 sterni (he/him) | In reply to @pwmosquito:matrix.org
not sure if anyone has any opinion on it but the active-repositories: none Cabal issue is being discussed here:
https://github.com/haskell/cabal/issues/10504
for reference i've used this for years to essentially tell Cabal "build with the packages i provide" or in other words separate building (with Cabal cause it's incremental) from package management (cause we use Nix for that)
it’s not really broken just unfortunate program flow which is similar to the static pkg config problem | 17:29:07 |
| sterni (he/him) | active-repositories: none is documented and still works you just have to have some kind of index downloaded unfortunately now | 17:29:40 |
| sterni (he/him) | bit of a trend with cabal-install that it’s over eager in checking requirements | 17:30:04 |
| sterni (he/him) | In reply to @lxsameer:matrix.org
do you have any suggestion on how to approach this problem with dependencies and tests? add mkDerivation = args: super.mkDerivation (args // { doCheck = false; }) | 17:31:08 |
| sterni (he/him) | then you won’t need to emit dontCheck as well | 17:31:40 |
| lxsameer | In reply to @sternenseemann:systemli.org
add mkDerivation = args: super.mkDerivation (args // { doCheck = false; }) hmmm, that seems scary to be honest. :D | 17:32:42 |
| sterni (he/him) | that’s the best way to disable test globally in the package set | 17:33:08 |
| sterni (he/him) | I think it’s even a documented pattern :) | 17:33:26 |
| lxsameer | I'll try it now | 17:34:05 |
| lxsameer | In reply to @sternenseemann:systemli.org
I think it’s even a documented pattern :) yupe that works. thanks a mil | 17:45:02 |
| pwmosquito | In reply to @sternenseemann:systemli.org
active-repositories: none is documented and still works you just have to have some kind of index downloaded unfortunately now It broke a bunch of build scripts at work that worked fine for years. Looks like they’ll revert these errors back to warnings for now which is good (cause atm we have to patch cabal-install). But long term i think we need a flag for explicitly saying “don’t manage my packages, only build with what i provide” | 17:47:27 |
| maralorn | In reply to @sternenseemann:systemli.org
add mkDerivation = args: super.mkDerivation (args // { doCheck = false; }) What I don't like about this is that it destroys cache reuse. | 17:52:19 |
| sterni (he/him) | it’s wuite unlikely that you’d have that anyways | 17:53:11 |
| sterni (he/him) | you should probably disable profiling as well while you’re at it…. | 17:53:44 |
| lxsameer | In reply to @sternenseemann:systemli.org
you should probably disable profiling as well while you’re at it…. is there a flag for that too? | 17:55:27 |
| lxsameer | In reply to @maralorn:maralorn.de
What I don't like about this is that it destroys cache reuse. only from the binary cache right? | 17:55:56 |
| sterni (he/him) | enableLibraryProfiling is on by default | 17:56:07 |
| lxsameer | In reply to @sternenseemann:systemli.org
enableLibraryProfiling is on by default by overriding the mkDerivation, user still would be able to override the derivation and enable checks or profiling, right? | 17:59:52 |
| sterni (he/him) | no, unfortunately not | 18:04:28 |
