Re: sbuild on hurd-amd64...

Samuel Thibault, le mer. 20 août 2025 02:48:22 +0200, a ecrit:
> Stéphane Glondu, le lun. 18 août 2025 08:20:49 +0200, a ecrit:
> > With the attached boot-script.hurd (taken from hurd's boot.c and adapted to
> > use /usr/lib/ld-x86-64.so.1) and a subhurd.img (as an ext2 image generated
> > using debootstrap), I get:
> > 
> > > demo@debian:~$ boot --boot-script=boot-script.hurd subhurd.img
> > > /hurd/ext2fs.static --readonly "--multiboot-command-line=boot -
> > > root=pseudo-root" --host-priv-port=1 --device-master-port=2
> > > --kernel-task=3 --exec-server-task=4 -T device pseudo-root
> > > /lib/ld-x86-64.so.1 /hurd/exec
> > 
> > Then nothing. I have no idea if something is going on, nor do I know how to
> > proceed to debug things. Any pointers?
> 
> boot was not actually ever tested on 64b, its ELF loading was assuming
> 32b. I have now fixed it upstream and uploaded a -3 package, which
> should get built & uploaded soonish.

It's now in.

> FI, apparently ATM when using a file as image, libstore uses 1-byte
> block size, and thus disk sizes should be smaller than 2G otherwise
> boot's device size reporting interface gets wrong.

(an alternative is to use a disk partition, which report 512-byte block
size)

Perhaps we could make libstore use 512-byte block size when the file is
multiple of 512 bytes. It could also be configurable, by introducing a
libstore layer (similarly to the part libstore layer)

Samuel