Source: dh-haskell Version: 0.6.11b Severity: minor User: [email protected] Usertags: toolchain X-Debbugs-Cc: [email protected]
Hi, I'm not at all sure whether this is actually dh-haskell's fault, but whilst working on the Reproducible Builds effort [0], I noticed that Haskell packages are building different binaries depending on whether "nocheck" is specified. (I'm filing it here because it seems like the first step.) For instance, when I build the haskell-integer-logarithms package and I run the testsuite (ie. I don't specify "nocheck"), I get the following: 283de3b00123c50ff754a7b7501e52efd9c8bdb9 libghc-integer-logarithms-dev_1.0.3.1-4_amd64.deb However, when I do specify "nocheck", I get the following: df5c67f1d255d04d4fcc803b30088fdd08d69481 libghc-integer-logarithms-dev_1.0.3.1-4_amd64.deb These results are determinstic — ie. I get the same two SHAs if I repeat the 'check' and nocheck builds. I wonder if this is an ABI hash issue; GHC is calculating a different hash because of the code/symbols of the testsuite part? https://tests.reproducible-builds.org/debian/dbdtxt/unstable/amd64/haskell-integer-logarithms_1.0.3.1-4.diffoscope.txt.gz Indeed, poking some more makes me think the issue is not with dh-haskell, which is AFAICT simply passing --enable-tests? Your help appreciated. :) Patch attached. [0] https://reproducible-builds.org/ Regards, -- ,''`. : :' : Chris Lamb `. `'` [email protected] / chris-lamb.co.uk `-
