I showed that `aptitude show libasyncns0` (with or without ':arch' or '=ver' qualifiers) shows 'Breaks:' and 'Replaces:' lines for the 'other' arch.
But -- `apt show` does not show those lines; `dpkg -I` on the *.deb doesn't show; and extracting the packages' 'control' files also doesn't show them! Yet, trying to actually test-install (with `apt-get -s install` or `apt -s install`) *does* report all the conflict consequences as implied by the `aptitude` output. Ahhh, I think I understand: it is a consequence of the 'Multi-Arch: same' header. And there doesn't seem to be a value for that header which means 'multiple arch versions can be installed at once, and it is OK if they aren't the same version'. I suspect this can be 'simulated' by the use of explicit 'Breaks' and 'Replaces' headers giving wider leeway; but that sounds like an extra bad maintenance nightmare... No, the answer really seems to be: do not issue updates to just one arch at once, of a Multi-Arch package. It just doesn't work. In fact it is likely enough to cause trouble, that I would kind of expect the maintainer tools (upload scripts etc) to try to call it out and prevent it... (It seems that `aptitude` is basically displaying 'imputed' 'Breaks' and 'Replaces' headers -- what the package *could* have explicitly included in its control file to get approximately the same effect as the 'Multi-Arch: same' header. This `aptitude` behavior was both confusing and, in the end, helpful and enlightening...)
