Summary of the issue
retitle 396122 kernel-package: fails when a kernel-patch adds a config var without a default value
The issue, as discussed on IRC, is that newer kbuild runs "make
silentconfig" very often, and that somewhat broke kernel-package
A possible fix would be:
- save .config as early as possible
- run "make defconfig" as second step, so any further make invocation is
guaranted not to fail on "silentconfig"
- restore the saved .config just before running "make <whatever>config"
OTOH, it may be worth to identify the failing "sed" - the problem is
The problem with this particular issue is that it makes unusable all kernel-patch
packages which add a config var without giving a default answer. And
"fixing" those to add a default answer will break the user's
expectation, as I suppose many do the same as I do: wait for "make
oldconfig" to present the choice, to decide what to activate.
This is why I propose this bug to be raised to "serious".
As a sidenote, some things mentionned on IRC that may be of interest:
- strace reports culprit as ["sed", "-e", "s%* * Restart config... * * * Pr"..., "oldconfig to update configuratio"..., ".mak"]
- ./.mak shows "Work around bad spots in RAM (BADRAM) [Y/n/?] (NEW) aborted!"