Checksums don't add security, that's why they're the "integrity check", not the "security check". Do you know how many [core] packages don't have PGP signatures available at all? Those are used on far more devices.
Granted, using PGP when available is always nice. But I don't see you screeching at the non-dkms package maintainer to fix his packages...
Edit: to clarify, I even like strong integrity checks myself, because they're definitely better than nothing and it can only help. But you're going about this totally the wrong way and you should also consider the old saying about people who live in glass houses.
Please add some kind of checksum checking to this package. Currently, the source integrity fully relies on a valid https certificate and the server behind it returning the right data. This doesn't sound right for a kernel module used in thousands of devices.
You can switch to a download link of the release, instead of a git clone (which also reduces the download time and the server load) like this:
https://github.com/zfsonlinux/spl/archive/spl-0.7.8.tar.gz
Then you can just add a checksum for this archive.
Could the aarch64 architecture be added to the PKGBUILD? The ZFS on Linux projects states that the arch is supported (https://github.com/zfsonlinux/zfs/wiki/FAQ) and I've been using a modified PKGBUILD compiled on aarch64 successfully for some time as well.
@risto3
At least until the packages are updated here do the following:
The 7.6 version of spl and zfs fixes the issue. when installing spl-dkms or zfs-dkms edit the PKGBUILD.
in both zfs-dkms and spl-dkms, Change the variable pkgver=0.7.5 to pkgver=0.7.6
continue building.. it should install properly.
Gulp, I just got the following upgrading today:
==> dkms install spl/0.7.5 -k 4.15.1-2-ARCH
Error! Bad return status for module build on kernel: 4.15.1-2-ARCH (x86_64)
Consult /var/lib/dkms/spl/0.7.5/build/make.log for more information.
==> dkms install acpi_call/1.1.0 -k 4.15.1-2-ARCH
==> WARNING: Cannot resolve dependencies for module zfs/0.7.5, kernel version 4.15.1-2-ARCH
(3/5) Updating linux initcpios...
the error is:
/var/lib/dkms/spl/0.7.5/build/module/spl/spl-taskq.c: Dans la fonction « taskq_dispatch »:
/var/lib/dkms/spl/0.7.5/build/module/spl/spl-taskq.c:593:16: error: « struct timer_list » n'a pas de membre nommé « data »
t->tqent_timer.data = 0;
^
/var/lib/dkms/spl/0.7.5/build/module/spl/spl-taskq.c: Dans la fonction « taskq_dispatch_delay »:
/var/lib/dkms/spl/0.7.5/build/module/spl/spl-taskq.c:643:16: error: « struct timer_list » n'a pas de membre nommé « data »
t->tqent_timer.data = (unsigned long)t;
^
/var/lib/dkms/spl/0.7.5/build/module/spl/spl-taskq.c:644:26: error: affectation depuis un type pointeur incompatible [-Werror=incompatible-pointer-types]
t->tqent_timer.function = task_expire;
^
CC [M] /var/lib/dkms/spl/0.7.5/build/module/splat/splat-condvar.o
/var/lib/dkms/spl/0.7.5/build/module/spl/spl-taskq.c: Dans la fonction « taskq_init_ent »:
/var/lib/dkms/spl/0.7.5/build/module/spl/spl-taskq.c:735:2: error: déclaration implicite de la fonction « init_timer »; vouliez-vous utiliser « init_timers » ? [-Werror=implicit-function-declaration]
init_timer(&t->tqent_timer);
^~~~~~~~~~
(excuse my French :-)
needs spl-0.7.6 for 4.15 patches
Simply replacing "0.7.3" with "0.7.4" in the PKGBUILD seems to work for me.
(When used with zfs-dkms, I did the same there and uninstalled both packages prior to reinstalling the newer versions.)
Hello! We just added support to the archzfs project and repository for these PKGBUILDS. Could we discuss adding the archzfs maintainers on as co-maintainers to this AUR package? See https://github.com/archzfs/archzfs/issues/157
The AUR package should be updated for 0.6.5.10 but in the meantime I have been able to successfully build and install spl-dkms & zfs-dkms 0.6.5.10 for linux 4.11.6. You need downgrade gcc to version <= 6.3.1 first. Download the snapshot for spl-dkms & zfs-dkms and edit the pkgver from 0.6.5.9 to 0.6.5.10 in the PKGBUILD files for each.
Checksums don't add security, that's why they're the "integrity check", not the "security check". Do you know how many [core] packages don't have PGP signatures available at all? Those are used on far more devices.
Granted, using PGP when available is always nice. But I don't see you screeching at the non-dkms package maintainer to fix his packages...
Edit: to clarify, I even like strong integrity checks myself, because they're definitely better than nothing and it can only help. But you're going about this totally the wrong way and you should also consider the old saying about people who live in glass houses.