<davidlt[m]>
Clang is cooked. Recompiling LLVM with Clang. I wonder which one is faster, GCC vs. Clang.
<davidlt[m]>
Our current LLVM stack was compiled with GCC (which is not default), but Clang driver was broken (not a huge problem).
<davidlt[m]>
So that seems to fixed now and cooking. So back to compiling LLVM, following by Clang, other related packages that want to use Clang as default toolchain (and that toolchain is also built with Clang).
<davidlt[m]>
I also want to update Rust in parallel, but that's basically the only bits that I want to finalize before a new Fedora disk image goes into virt-builder repo.
<davidlt[m]>
Well, maybe a few other things between all of this.
<davidlt[m]>
Back to experimenting with Btrfs.
<davidlt[m]>
I have not idea why I decided to use Gnome Boxes instead of libvirt.
<davidlt[m]>
Another thing why I am looking into Btrfs as new /mnt/koji is transparent compression support.
<davidlt[m]>
I also like the fact that I can add new devices, which can be of a different size. If we run out of the disk space again.
<davidlt[m]>
So with zstd:1 GCC build.log (~46M) it gets down to ~13M in size.
<davidlt[m]>
One thing that we do different from the upstream Koji (well, a lot), but in general we don't prune failed tasks logs.
<davidlt[m]>
I do like to pull those logs out and figure out if things have changed over the years, or what was the issue in the 1st place.
<davidlt[m]>
So yeah, I do pull logs from 2018. Sometimes I still go and look into the logs before we even had Koji running.
<davidlt[m]>
Oh, Linux patches reveal that next-gen AMD APUs (laptops at least) will have RDNA3 iGPU it seems.
<davidlt[m]>
rmlint seems to be nice to find duplicate files
<davidlt[m]>
seems Btrfs doesn't have inband block level dedup (not that it would be a problem for us)
<davidlt[m]>
So in the current Koji setup it seems that /mnt/koji/work has only 5.90G of duplicates (files wise).
<davidlt[m]>
I did find a bit more free space too from cli-build directory. Those could be garbage collected without no issues.
<davidlt[m]>
Other than that tasks is the the largest contributors to disk usage.
davidlt has quit [Ping timeout: 264 seconds]
davidlt has joined #fedora-riscv
<davidlt[m]>
I think, it's time to move experiments into AWS new Koji server :)
zsun has quit [Quit: Leaving.]
jcajka has quit [Quit: Leaving]
<davidlt[m]>
I get impression that Clang is compiling faster, but TBD.
<davidlt[m]>
New stable kernels coming in. 4.9.334 was tagged 1 minutes ago.
<davidlt[m]>
6.0.10 should land within hours too (most likely). Quite a large one, 313 patches the last time I checked.
esv_ is now known as esv
<davidlt[m]>
rwmjones: what did we use to execute a bunch of DNF commands in a single transaction? dnfdragora ?
<rwmjones>
err, I've never done it with dnf
<rwmjones>
but with yum there was a shell you could use
Kevinsadminaccou has joined #fedora-riscv
<rwmjones>
& you could run all the commands together in a single transaction
<rwmjones>
"dnf shell" exists, so ...
<davidlt[m]>
ah, that's what it is
<rwmjones>
seems to support transactions
<davidlt[m]>
That's the one I was looking for. So rarely used (at least by me).