| 1 |
* The dependancies between targets in the Makefile really need a lot of |
* Add disk labels to non-i386 floppys. |
| 2 |
work. |
* The busybox in the bootfloppy can be linked with uclibc and/or statically |
| 3 |
* The apt-get autoclean call isn't good enough; sometimes the cache still |
to make it smaller. Only the linuxrc uses it, so it doesn't really |
| 4 |
contains multiple versions of a udeb, which breaks the makefile. |
matter. This could save up to 200k on the boot floppy. But first someone |
| 5 |
|
needs to add code to install uclibc or a uclibc udeb. |
| 6 |
|
* Find something to do with spare space on the bootfloppy; put udebs there |
| 7 |
|
for extra drivers or something. |
| 8 |
|
* The apt download fails if the dpkg status file of the build system is |
| 9 |
|
in an inconsistent state. It could be overridden to use another status |
| 10 |
|
file, but that file would have to claim that all dependancies of the |
| 11 |
|
udebs it's downloading were met. Hmm. |
| 12 |
|
* The extra driver disks stuff is currently hardcoded for one disk, and |
| 13 |
|
should be generalised and generally cleaned up. |
| 14 |
|
* The size calcs in make stats for driver1 are not right, it can be full |
| 15 |
|
and will show as having space free. |
| 16 |
|
* Change mklibs to not pull in libraries from the build system, and only |
| 17 |
|
act on libraries in the tree; those libraries would then be provided by |
| 18 |
|
udebs; this would be cleaner in several ways (and it's why eg anna |
| 19 |
|
depends on libc-udeb already). |
| 20 |
|
* Switch alpha, powerpc, over to new "di" KERNEL_FLAVOUR. |
| 21 |
|
* Convert the driver floppies to use FAT filesystem, not ext2, on i386. |
| 22 |
|
Makes it easier for third parties to throw additional udebs on. |
| 23 |
|
* Cache the upxed kernel, to avoid re-upxing it over and over. |