Package: release.debian.org Severity: normal User: release.debian....@packages.debian.org Usertags: unblock
Due to a misconfiguration of ghc, where it uses the 'arm1136jf-s' ARM11 core family on armel, all Haskell packages are currently miscompiled and will only work on a subset of armel machines (the ones that use the ARMv6 architecture or newer). This has been reported as #915333. [#915333] https://bugs.debian.org/915333 In order to fix the above bug for buster, we have to: 1) Upload ghc/8.4.4+dfsg1-3 to unstable with the following fix: diff --git a/p/ghc/debian/patches/llvm-arm-unknown-linux-gnueabi.patch b/p/ghc/debian/patches/llvm-arm-unknown-linux-gnueabi.patch index 10fe15d0e4..0ac3a43a64 100644 --- a/p/ghc/debian/patches/llvm-arm-unknown-linux-gnueabi.patch +++ b/p/ghc/debian/patches/llvm-arm-unknown-linux-gnueabi.patch @@ -8,7 +8,7 @@ Index: b/llvm-targets ,("x86_64-unknown-windows", ("e-m:w-i64:64-f80:128-n8:16:32:64-S128", "x86-64", "")) ,("arm-unknown-linux-gnueabihf", ("e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64", "arm1176jzf-s", "+strict-align")) ,("armv6-unknown-linux-gnueabihf", ("e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64", "arm1136jf-s", "+strict-align")) -+,("arm-unknown-linux-gnueabi", ("e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64", "arm1136jf-s", "+strict-align")) ++,("arm-unknown-linux-gnueabi", ("e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64", "arm7tdmi", "+soft-float -fp-only-sp -d16 -vfp2 -vfp3 -fp16 -vfp4 -fp-armv8 -neon -crypto +strict-align")) ,("armv6l-unknown-linux-gnueabihf", ("e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64", "arm1176jzf-s", "+strict-align")) ,("armv7-unknown-linux-gnueabihf", ("e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64", "generic", "")) ,("armv7a-unknown-linux-gnueabi", ("e-m:e-p:32:32-i64:64-v128:64:128-a:0:32-n32-S64", "generic", "")) 2) binNMU on armel, all packages that were built with ghc 3) unblock ghc and binNMU-ed packages Please note that I do not have the means to verify that the above fix is working (other than asking the bug reporter to test git-annex after step 2 has been completed), since our armel buildds/porterboxes are based on ARM v7 (this is why our tests run successfully on buildds and we didn't catch this earlier). How should we proceed? Do you think this is something worth fixing before releasing buster? -- Ilias
