Hi, we have been disagreeing a bit about who should provide linux kernel headers for cross compilation until Brest and in that discussion the argument that them going out of sync could cause problems, but we never had any practical evidence of that happening.
Enter input-utils. It started failing lately https://crossqa.debian.net/src/input-utils, but what looks like a regression of an upload actually is pure coincidence with a concurrent linux upload. The package contains a name.sh that parses macros from /usr/include/linux/input-event-codes.h using awk and then emits tables. Those tables are then compiled using a (cross) compiler that happens to use c-t-b's older version of the file and lacks the #define of SW_USB_INSERT. It is not clear to me whether this is a sensible use, but it practically demonstrates the claim that the desync may cause problems. We have since agreed on letting linux provide the packages, so this issue will resolve itself before too long and all this message does is document another reason. We may also conclude that this kind of problem is exceptional. Helmut
