On Mon, 08 Jun 2015, Nikos Mavrogiannopoulos wrote:
On Mon, 2015-06-08 at 09:39 +0200, Dodji Seketeli wrote:
Of course, nobody likes ABI *breakage*. And I agree that if all ABI
breakage could be detected automatically, ABI breakages would never
make
it into stable releases.
The thing is, the tool detects ABI *changes*. Some changes are
breakages. Some are not, depending on the particular context we are
looking at. The tool does have heuristics to categorize certain
changes
as being ABI breakages, but ultimately, I believe there is going to
be
cases where human intervention is going to be necessary to tell if a
given change is harmful or not. And it's going to be so for the
foreseeable future. You can think of it as a kind of patch review,
but for ABI changes specifically.
I have not seen the output of abicheck (I use abi-compliance-checker
personally but I guess abidiff is as good). However, I'm not sure about
which changes which are not breakages you mean? I'm not aware of ABI
changes which do not break users of libraries.
Adding new functions to ABI constitute changes that don't break existing
users as long as previously available data structures are not affected.
--
/ Alexander Bokovoy
--
devel mailing list
devel@lists.fedoraproject.org
https://admin.fedoraproject.org/mailman/listinfo/devel
Fedora Code of Conduct: http://fedoraproject.org/code-of-conduct
