On Sat, 2013-04-13 at 13:15 +0100, Adam D. Barratt wrote: > Control: tags -1 + moreinfo > > On Sat, 2013-03-23 at 12:30 +0000, Alan Woodland wrote: > > Most of the patches that Debian was running in 0.8.2-X are now merged > > in upstream. Primarily the changes in this diff are adding support for > > newer kernels. There are some test suite updates and userspace library > > fixes, but no ABI breaking changes. I've also addressed a few Debian > > specific issues. > > > > The new kernel support has been extensively tested outside of > > Debian[1] and the process also uncovered and fixed a number of > > existing bugs[2]. > > > > I'm confident that this is the right version to release with. > > A couple of comments, having read back through the earlier discussion > regarding the proposal to remove the package. > > In <CAD2JkJeRD5fZHdiL1tmc4OqTM=gdGt=NN+u6OqffuTpUramu9g@mail.gmail.com>, > you indicated that "[t]he library behaves sanely without the kernel > module". If that's the case then I'm a little dubious about the merits > of re-adding the module package at this late stage of the freeze. > > One of our kernel maintainers indicated in that earlier discussion that > he doesn't believe the c/r functionality should be in an out-of-tree > module. Ben, I assume that's still the case? What I meant was that checkpoint/resume necessarily depends on kernel interfaces that are not particularly stable and not designed to provide all the information needed, so the breakage of blcr kernel modules can be expected to happen again and to be difficult to fix. But I don't want to stand in the way of an update to blcr that makes it work out of the box in Debian 7.0. It will provide at least a stopgap until people can use CRIU (checkpoint/restore in userspace) which I expect to be more or less complete in Debian 8.0. Ben. -- Ben Hutchings Humans are not rational beings; they are rationalising beings.
Attachment:
signature.asc
Description: This is a digitally signed message part