Tests fail on big-endian systems

[iliastsi]

Tests fails on big-endian systems. Here is an example:

  test/Data/StaticBytesSpec.hs:65:20: 
  14) Data.StaticBytes, Primitive Vector, handles bytes16
       Falsifiable (after 4 tests and 3 shrinks):
         [1]
       expected: [1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
        but got: [0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0]

  To rerun use: --match "/Data.StaticBytes/Primitive Vector/handles bytes16/" --seed 1014189549

  test/Data/StaticBytesSpec.hs:75:20: 
  15) Data.StaticBytes, Primitive Vector, handles bytes32
       Falsifiable (after 4 tests and 3 shrinks):
         [1]
       expected: [1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
        but got: [0,0,0,0,0,0,0,1,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]

  To rerun use: --match "/Data.StaticBytes/Primitive Vector/handles bytes32/" --seed 1014189549

  test/Data/StaticBytesSpec.hs:83:21: 
  16) Data.StaticBytes, Primitive Vector, fuzz with encodeUtf8
       Falsifiable (after 2 tests and 1 shrink):
         "a"
       expected: [97,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]
        but got: [0,0,0,0,0,0,0,97,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0]

  To rerun use: --match "/Data.StaticBytes/Primitive Vector/fuzz with encodeUtf8/" --seed 1014189549

For the full build log see https://buildd.debian.org/status/fetch.php?pkg=haskell-static-bytes&arch=s390x&ver=0.1.0-1&stamp=1729262862&raw=0

[mpilgrem]

@iliastsi, thanks for reporting. In case the build log is ephemeral, I'm posting more from it below:

Data.StaticBytes
  ByteString
    padding is the same as trailing nulls [✘]
    handles bytes16 [✘]
    handles bytes32 [✘]
    fuzz with encodeUtf8 [✘]
  Storable Vector
    padding is the same as trailing nulls [✘]
    handles bytes16 [✘]
    handles bytes32 [✘]
    fuzz with encodeUtf8 [✘]
  Unboxed Vector
    padding is the same as trailing nulls [✘]
    handles bytes16 [✘]
    handles bytes32 [✘]
    fuzz with encodeUtf8 [✘]
  Primitive Vector
    padding is the same as trailing nulls [✘]
    handles bytes16 [✘]
    handles bytes32 [✘]
    fuzz with encodeUtf8 [✘]

and

test/Data/StaticBytesSpec.hs:60:47: 
  13) Data.StaticBytes, Primitive Vector, padding is the same as trailing nulls
       Falsifiable (after 1 test and 2 shrinks):
         (0,0,1,0)
       expected: Right "\NUL\NUL\SOH\NUL\NUL\NUL\NUL\NUL"
        but got: Right "\NUL\NUL\NUL\NUL\NUL\SOH\NUL\NUL"

  To rerun use: --match "/Data.StaticBytes/Primitive Vector/padding is the same as trailing nulls/" --seed 1014189549

[mpilgrem]

As an accident of its history (spun out of Stack via pantry), I think this package implicitly assumes the architecture, and Word64 values, are little endian. Possibly because the GHC project only distributes GHC binaries for little endian architectures (x86_64, AArch64).

[iliastsi]

As an accident of its history (spun out of Stack via pantry), I think this package implicitly assumes the architecture, and Word64 values, are little endian. Possibly because the GHC project only distributes GHC binaries for little endian architectures (x86_64, AArch64).

I see that in Debian we have disabled the tests for pantry, with a comment that they try to access the network (which is not allowed). This is probably why we failed to detect this earlier, and we have shipped (possibly broken) Stack on big-endian systems so far. For now, I will remove Stack from those systems.

[mpilgrem]

@iliastsi, the version of the package in branch fix1 is intended to fix the problem, but I have no means to test it on a big-endian machine architecture. Would you be able to do so?

[iliastsi]

@mpilgrem PR #4 solves the issue. Tests are now passing on big-endian systems. See here for the build logs https://buildd.debian.org/status/fetch.php?pkg=haskell-static-bytes&arch=s390x&ver=0.1.0-3&stamp=1729513039&raw=0.

Thank you for fixing this so quickly.