Comment 5 for bug 2015867

Just picking on linux-headers for now it is more or less what I expected.

It went 26 -> 107 mb and I see new directories for rust of:
81M linux-headers-6.2.0-1003-kvm_6.2.0-1003.3_amd64.deb.extract/usr/src/linux-headers-6.2.0-1003-kvm/rust
228K linux-headers-6.2.0-1003-kvm_6.2.0-1003.3_amd64.deb.extract/usr/src/linux-headers-6.2.0-1003-kvm/include/generated

I know linux-headers are meant to be the full source, but this increases the size of a default installed (and therefore also often downloaded) package by x4.
You see that even the compressed size increased, actually even more than extracted.

I'm not sure we can, but could we think about separating that and not installing by default.
I'm afraid that might break things but it juts feels rather "expensive" in terms of disk space, image size, regular download ...

OTOH, looking at it maybe we can do something...
I know rust is using static linking and it looks like here also a lot pre-generated content.

/tmp/kernel-sizes/6.2/linux-headers-6.2.0-1003-kvm_6.2.0-1003.3_amd64.deb.extract/usr/src/linux-headers-6.2.0-1003-kvm/rust:

$ ll -S | head
total 81876
-rw-r--r-- 1 paelzer paelzer 54727271 Apr 6 09:50 libcore.rmeta
-rw-r--r-- 1 paelzer paelzer 18678432 Apr 6 09:50 libmacros.so
-rw-r--r-- 1 paelzer paelzer 5380859 Apr 6 09:50 libbindings.rmeta
-rw-r--r-- 1 paelzer paelzer 2835696 Apr 6 09:50 core.o
-rw-r--r-- 1 paelzer paelzer 904778 Apr 6 09:50 liballoc.rmeta
-rw-r--r-- 1 paelzer paelzer 589288 Apr 6 09:50 bindings.o
-rw-r--r-- 1 paelzer paelzer 186221 Apr 6 09:50 libkernel.rmeta
-rw-r--r-- 1 paelzer paelzer 106528 Apr 6 09:50 kernel.o
-rw-r--r-- 1 paelzer paelzer 99356 Apr 6 09:50 .bindings.o.cmd

Since these are meant to be sources, could we provide them in a way to be regeneratable but not of the binary size and format in linux-headers?