with the recent ABI change, it was non-obvious, which "ABI" part
changed, so I would like to propose to provide a little bit more of
information, where the ABI is changing.

Please consider shipping one file together with the .ali files, e.g.
ali-checksums.xz, which lists per line
<name of ali file>  <xor'ed checksums from that file>

and then calculate the gnat checksum from that file.  That way, it's
easier to get a hint on what maybe is causing an ABI change for the gnat
checksum.

Hello.

I once had the same problem and spent a hard time comparing the old
and new .ali files.
The solution was to forget them once they have played their role of
alert, and compare the sources directories with something like 'diff
-urN'.
When we search for an unexpected ALI breakage, the changes are
unexpected, hence usually quite small.