zsun has joined #fedora-riscv
zsun has quit [Quit: Leaving.]
davidlt has joined #fedora-riscv
<
davidlt>
rwmjones_, OMG, GHC is still writing out RPMS
<
davidlt>
it spent the whole night at ~65 load average writing out RPMs
zsun has joined #fedora-riscv
<
davidlt>
ghc build just landed
<
davidlt>
let's try to use mock to test it
zsun has quit [Quit: Leaving.]
<
davidlt>
rwmjones_, we are good, time for Haskell mass rebuild :)
<
davidlt>
rwmjones_, nufive.home.annexia.org is off
<
rwmjones_>
davidlt: pandoc built
<
rwmjones_>
let me restart nufive
<
rwmjones_>
I restarted kojid on nufive so hopefully it's up now
<
rwmjones_>
but it's rather hard unless I manually download and examine every single git repo
<
davidlt>
rwmjones_, did you see any email about branching? I see branches, but not that it's done.
<
rwmjones_>
I see branches too, but didn't look for emails about it
<
rwmjones_>
any suggestions on ^^ ?
<
rwmjones_>
it would be nice to have a list of packages that have downstream changes (except just bumping the release)
<
davidlt>
the way you check is: koji list-tagged --latest --quiet f40 | grep '\.riscv64'
<
davidlt>
all modified packages incl.: .X.riscv64%{?dist}
<
davidlt>
So that information is encoded in NVR (and thus we have unique one).
<
davidlt>
The original NVR from upstream Koji does not get used.
<
rwmjones_>
but for example ghc is 0.riscv64, but the only change is a release bump
<
davidlt>
Yes, because there is no upstream NVR with those changes.
<
davidlt>
I am trying to be careful with NVRs.
<
rwmjones_>
ok I see
<
davidlt>
They can only be used __once__.
<
davidlt>
rwmjones_, you want to start sending easy ones?
<
davidlt>
99% of valgrind ones are literally 1-3 line changes.
<
davidlt>
These should be quite a good number of them.
<
rwmjones_>
I want to help as much as possible getting changes into Fedora dist-git
<
davidlt>
some others require retesting, or a bit more proper change.
<
rwmjones_>
by "valgrind" you mean changes to other packages that need to disable valgrind or similar?
<
davidlt>
zlib-ng, and gdal should be easy ones too.
<
davidlt>
Nah, majority of package aren't properly checking valgrind support for arch.
<
davidlt>
Instead they use random list of arches, or just assume it's everywhere.
<
davidlt>
Instead they need to check: %ifarch %{valgrind_arches}
<
davidlt>
It's basically just a clean up to properly check %{valgrind_arches} before doing anything valgrind related
<
davidlt>
In most cases this is just for BR: valgrind, but in a few cases it also needs checks in %check or/and %build to disable it
<
davidlt>
These would be a quick and easy win, probably ~20 (maybe more) packages.
<
rwmjones_>
using valgrind_arches is just good practice
<
davidlt>
isomd5sum is an easy win too (not valgrind related)
<
davidlt>
rwmjones_, another "quick win" is to ask GHC to get a proper upstream Koji build :)
<
davidlt>
This way I could remove .X.riscv64 build
<
rwmjones_>
yeah I'll email him now
<
rwmjones_>
keep dropping package names in here that have changes, and I'll make PRs for them
<
rwmjones_>
so far I'm looking at: zlib, gdal, jemalloc, isomd5sum
<
davidlt>
there is no zlib anymore IIRC
<
davidlt>
libunwind should be trivial too
<
rwmjones_>
zlib-ng I meant!
<
davidlt>
volk is trivial too
<
davidlt>
be back in a bit, time to get some food
davidlt has quit [Ping timeout: 255 seconds]
davidlt has joined #fedora-riscv
<
rwmjones_>
davidlt: ok I did those packages
<
davidlt>
all of them?
<
rwmjones_>
all the ones you mentioned, yes
<
rwmjones_>
I've CC you on all the pull requests
zsun has joined #fedora-riscv
<
rwmjones_>
installing texlive packages is painful, the %post script does a huge amount of random stuff
<
davidlt>
ok, I will make another short list
<
davidlt>
this might require sending a patch upstream if it's not done yet
<
davidlt>
otherwise trivial
<
rwmjones_>
ok, I'll get started on that lot in a min
<
rwmjones_>
davidlt: btw the volk package doesn't actually build for me on riscv64:
<
rwmjones_>
-- cpu_features package not found. Requiring cpu_features submodule ...
<
rwmjones_>
CMake Error at CMakeLists.txt:127 (message):
<
rwmjones_>
cpu_features/CMakeLists.txt not found. Did you forget to git clone
<
rwmjones_>
recursively?
<
rwmjones_>
Fix with: git submodule update --init
<
davidlt>
rwmjones_, maybe version got updated?
<
davidlt>
well, looknig at the spec it's done for s390x
<
davidlt>
should work
<
davidlt>
and it worked fine in volk-3.1.1-1.fc40 14 days ago
jednorozec has joined #fedora-riscv
davidlt has quit [Ping timeout: 255 seconds]
zsun has quit [Remote host closed the connection]
skip77 has quit [Ping timeout: 256 seconds]
skip77 has joined #fedora-riscv
zsun has joined #fedora-riscv
zsun has quit [Client Quit]
zsun has joined #fedora-riscv
zsun has quit [Quit: Leaving.]
zsun has joined #fedora-riscv
davidlt has joined #fedora-riscv
zsun has quit [Quit: Leaving.]
<
rwmjones_>
davidlt: re libcap, the upstream spec file is now using %{golang_arches}
<
rwmjones_>
so I don't think we need to do anything there
<
rwmjones_>
(because %{golang_arches} includes riscv64)
<
davidlt>
cool, they updated :)
<
davidlt>
rwmjones_, I think Nick is working on landing binutils in Fedora 41/Rawhide now
<
rwmjones_>
davidlt: ok I've done all those packages now
<
rwmjones_>
You should be CC'd on everything I hope
<
rwmjones_>
if you have any more packages just drop them in here
<
davidlt>
a bit later
davidlt has quit [Remote host closed the connection]
davidlt has joined #fedora-riscv
<
rwmjones_>
davidlt: I forgot to mention, there's a question that you might be able to answer here:
<
davidlt>
rwmjones_, done
<
davidlt>
rwmjones_, anyways zlib-ng doesn't detect vectors dynamically
<
davidlt>
I wonder if this works at least on code level otherwise it will be quite some time before we enable it
cyberpear has joined #fedora-riscv
davidlt has quit [Ping timeout: 264 seconds]
davidlt has joined #fedora-riscv
davidlt has quit [Ping timeout: 264 seconds]
cyberpear has quit [Quit: Connection closed for inactivity]