Hi Tomi--

On Sat 2020-05-02 00:15:57 +0300, Tomi Ollila wrote:
> I did not see anything suspicious in code, but
>
> I got these test failures:
>
> in ubuntu 19.10 native environment, and
>
> in debian 10 (podman) container running in fedora 31 system
>
>
> T355-smime: Testing S/MIME signature verification and decryption
>  FAIL   Verify signature on PKCS#7 SignedData message
>  crypto: value not equal: data[0][0][0]["crypto"]["signed"]["status"][0] =
>  {'status': 'good', 
>   'fingerprint': '702BA4B157F1E2B7D16B0C6A5FFC8A7DE2057DEB', 
>   'created': 1574813489,
>   'expires': 2611032858} != 
>  {'created': 1574813489, 
>   'expires': 2611032858,
>   'fingerprint': '702BA4B157F1E2B7D16B0C6A5FFC8A7DE2057DEB', 
>   'userid': 'CN=Alice Lovelace', 
>   'status': 'good'}
>
> T356-protected-headers: Testing Message decryption with protected headers
>  FAIL   verify signed PKCS#7 subject (multipart-signed)
>  sig_uid: object not found:  data[0][0][0]["crypto"]["signed"]["status"][0]["userid"]
>  FAIL   verify signed PKCS#7 subject (onepart-signed)
>  sig_uid: object not found: data[0][0][0]["crypto"]["signed"]["status"][0]["userid"]
>  FAIL   confirm signed and encrypted PKCS#7 subject (sign+enc)
>  sig_uid: object not found: data[0][0][0]["crypto"]["signed"]["status"][0]["userid"]
>  FAIL   confirm signed and encrypted PKCS#7 subject (sign+enc+legacy-disp)
>  sig_uid: object not found: data[0][0][0]["crypto"]["signed"]["status"][0]["userid"]

Thanks for identifying these.  These are problems related to a bug in
the released version of GMime on those platforms.  Unfixed versions of
gmime cannot report *any* certificate validity for X.509 certificates:

   https://github.com/jstedfast/gmime/pull/90

The fix for gmime is pretty simple, but it's not something we can
address directly in notmuch.

The fix was first released in GMime version 3.2.7, but it was first in
debian in gmime 3.2.6-2, and should be relatively easy to backport for
any distro that wants it (i suppose i could probably get it into the
next point release for debian 10 as well, since it is a bugfix for an
already-exposed API).

So, how should we deal with this in notmuch?  It seems a bit silly to
bump our required version of gmime to the (relatively new) version
3.2.7, for a fix for a cornercase of a novel use case.

Maybe the test suite should change based on version of GMime?  That
would cause problems for distros that backport the GMime fix, though.

I guess i could write a reproducer for the gmime issue and we could
include it in ./configure, and modify the test suite on that basis.

Any other suggestions?

    --dkg