Calibre test suite fails

Hi,

Both on my local machine and on the CI[0], the Calibre test suite has

been failing with the following error:

======================================================================

ERROR: test_msgpack (calibre.test_build.BuildTest)

----------------------------------------------------------------------

Traceback (most recent call last):

File "/tmp/guix-build-calibre-3.42.0.drv-0/calibre-3.42.0/src/calibre/test_build.py", line 124, in test_msgpack

self.assertEqual(obj, msgpack_loads(s))

File "/tmp/guix-build-calibre-3.42.0.drv-0/calibre-3.42.0/src/calibre/utils/serialize.py", line 113, in msgpack_loads

return msgpack.unpackb(dump, ext_hook=msgpack_decoder, raw=False, use_list=use_list)

File "/gnu/store/z7dz4iiaivmadhk0x50qs5zv3rwykrmd-python2-msgpack-1.0.0/lib/python2.7/site-packages/msgpack/fallback.py", line 129, in unpackb

ret = unpacker._unpack()

File "/gnu/store/z7dz4iiaivmadhk0x50qs5zv3rwykrmd-python2-msgpack-1.0.0/lib/python2.7/site-packages/msgpack/fallback.py", line 666, in _unpack

"%s is not allowed for map key" % str(type(key))

ValueError: <type 'int'> is not allowed for map key

----------------------------------------------------------------------

This is happening on both staging and master. There have been no

changes to the calibre package itself since January. I have run a git

bisect and the offending change appears to be in commit

66ab2f5e3a0df665f6e39203aedd2bf4812e6a71, when python-msgpack was

updated to version 1.0.0.

[0]: e.g. http://ci.guix.gnu.org/build/2485977/details

--

Josh Holland

I've reported this bug as

https://bugs.launchpad.net/calibre/+bug/1870541. I also found the

backwards-incompatible change in msgpack which is the culprit of this:

hashdos. You need to pass strict_map_key=False if you have data which

contain map keys which type is not bytes or str.

(from https://pypi.org/project/msgpack/)

I can prepare a patch (both within the Guix build system and to send

upstream) which fixes this, though I'm not sure what hashdos are or

whether they are actually important to avoid in this case.

--

Josh Holland

Josh Holland <josh@inv.alid.pw> writes:

Thanks for reporting it! I was just looking into the issue and found

that it was fixed upstream 3 days ago:

I adopted the patch to Guix in 3e4ce1cc3eb7b35405c4d2c4f837e53ec9952c99.