On Fri, 20 Mar 2015, Ben Hutchings wrote: > On Fri, 2015-03-20 at 19:48 -0400, Benjamin Kaduk wrote: > > > > I have only become the openafs maintainer relatively recently, so my > > apologies if this is well-trodden ground, but what is supposed to happen > > when the stable KPI changes? Is it really just that all consumers are > > supposed to scramble to update their packages? > > Yes, I'm afraid so.
Sigh. Thanks for confirming it, at least. > > P.S. If I understand correctly, the issue with d_alias is that d_rcu was > > previously overlapping with d_child in a union, and that was > > incorrect/buggy. There was no need to move it to overlapping with d_alias > > other than to save a bit of space; the KBI was already changing, so the > > KPI could have been preserved by just moving d_child out of the union and > > accepting the extra space as the price of preserving the KPI. > > The ABI didn't change in Debian's kernel packages. Modules never use > d_rcu and none of the other fields moved. genksyms would generate > different symbol versions all over the kernel, but I hid this change to > prevent that happening. Ah, thank you for that... > So the openafs modules already built should continue to work, but can't > be rebuilt against the new headers. ...and that would explain why there are not already huge outcries coming my way :) Thanks again, Ben -- To UNSUBSCRIBE, email to [email protected] with a subject of "unsubscribe". Trouble? Contact [email protected]
