Hi,
I just was surprised (not to say a bit of shocked) by the libreoffice 4:24.2.1~rc2-1 build in experimental succeeding on riscv64.
(Since if this happens in unstable we get a build and then a subsequent FTBFS due to the test (see below) will be RC and blocking...)
As already discusses on -riscv64 libreoffice needs NaN payload passing and there is an upstream unit test testing for this.
Which resulted in
test-rtl-math.cxx:662:Assertion
Test name: (anonymous namespace)::Test::test_payloadNaN
equality assertion failed
- Expected: 48879
- Actual : 0
- Your platform does not support propagation of NaN payloads.
Failures !!!
Run: 1335 Failure total: 1 Failures: 1 Errors: 0
so far. (see e.g.
https://buildd.debian.org/status/fetch.php?pkg=libreoffice&arch=riscv64&ver=4%3A24.2.0-1&stamp=1706791493&raw=0 from rv-osuosl-03)
For buildd/admin people: See
https://lists.debian.org/debian-riscv/2024/01/msg00018.html and
https://bugs.documentfoundation.org/show_bug.cgi?id=152943 AFAICR correctly this was also the case when I tried a local build on my Milk V Mars board with -O2.
Now it succeeded on rv-osuosl-05 on the last upload to experimental.
There no change making that obsolete in 24.2.1 and I somehow don't believe -O2 would affect it?
Is rv-osuosl-05 hardware which supports this? (db.debian.org/machines.cgi doesn't really shed any light here; they all say "Hifive Unmatched").
Just running a build on my machine again, too.
Regards,
Rene