build boot1.efi for i386 as well

[emaste]

Summary

Loader still needs more work but boot1 builds now.
This builds on D2241.

[emaste]

sys/boot/efi/boot1/Makefile

43 ↗

(On Diff #4705)

i386 needs it for __divdi3, will either remove the explicit fn from the comment, or list both arm and i386 examples

sys/boot/efi/boot1/generate-fat.sh

23 ↗

(On Diff #4705)

yay consistency

[rpaulo]

sys/boot/efi/boot1/Makefile

43 ↗

(On Diff #4705)

Isn't that just because we're compiling i386 with -msoft-float? We probably don't need to do that for i386.

[imp]

sys/boot/efi/boot1/fat-i386.tmpl.bz2.uu

1 ↗

(On Diff #4705)

Why do we need these?

[emaste]

sys/boot/efi/boot1/Makefile

43 ↗

(On Diff #4705)

__divdi3 is 64 bit by 64 bit integer division, not fp.

sys/boot/efi/boot1/fat-i386.tmpl.bz2.uu

1 ↗

(On Diff #4705)

We need a FAT file system with boot1.efi in it, with a specified file name. We could do something a bit more fancy (and error-prone) than what we've got here, but the eventual plan is to add FAT support to makefs.

We'll probably only have four of these (i386, amd64, arm, arm64), so I'm not particularly interested in trying to come up with a hacky way to avoid having these.

[imp]

sys/boot/efi/boot1/fat-i386.tmpl.bz2.uu

1 ↗

(On Diff #4705)

If we can get makefs to generate these, we'll be better off. Though we'll need to address the date/time issue for repeatable builds. Having these files in the tree goes rather against checking in binaries that could be built. lack of support in makefs is a good reason though.

[emaste]

In D2244#8, @imp wrote:

If we can get makefs to generate these, we'll be better off. Though we'll need to address the date/time issue for repeatable builds. Having these files in the tree goes rather against checking in binaries that could be built. lack of support in makefs is a good reason though.

Right, if we didn't want to check these in we'd need to have buildworld run as root and invoke mdconfig.

[emaste]

Closed by commit rS281238 (authored by @emaste).