Rex Dieter wrote:
Pierre-Yves Chibon wrote:
> 1. dist.abicheck - to make sure the update's ABI remains stable in a
> given Fedora release.
I think it unwise to make item 1 a mandatory item at this point. I'd
argue a large number of packages do not provide public api/abi that's
worth worrying about in this regard.
+1
See:
https://pagure.io/fesco/issue/1810#comment-488673
https://lists.fedoraproject.org/archives/list/devel@lists.fedoraproject.o...
I wrote there:
Uh, `dist.abicheck` produces a lot of false positives on:
* libraries that are internal and that nothing should depend on (e.g., in
QupZilla, package `qupzilla`),
* APIs explicitly documented as "private, can change at any version", as
common in all Qt modules (e.g., in QtWebEngine, package
`qt5-qtwebengine`).
My packages often fail `dist.abicheck`. It is absolutely not realistic to
expect it to pass for all updates.
I don't think gating on this test will EVER be a reasonable thing to do. It
has just too many false positives. There is no automated way to figure out
which ABIs are intended to be public and which aren't. And even an API
"public" at package level can be private at distro level. (Consider a
library with exactly one package using it. It is always reasonable to bump
those together in a grouped update. We have several such examples in
Fedora.)
Adam Williamson wrote:
Additionally, as Ralph wrote, he has removed abicheck from the list
of
gating tests for now, so you shouldn't need to worry about abicheck
failures blocking updates, as soon as Bodhi starts applying that
change.
FYI, it looks like this is already working. (The other thread said it'd take
only up to 6 hours for Bodhi to pick it up, and indeed, qt5-qtwebengine is
now showing as passing the gating even though it "fails" the bogus abicheck
test.)
But why was this check enabled to begin with, when the issues were already
known (see the above links)? I get the feeling that I am talking to a wall!
Kevin Kofler