On Fri, Oct 04, 2024 at 02:59:06PM +0200, Santiago Vila wrote:
> Package: src:mlucas
> Version: 20.1.1-1.2
> Severity: serious
> Tags: ftbfs
>
> Dear maintainer:
>
> (I'm adding Bill to X-debbugs-Cc, since he did the last NMU)
Thanks!
> During a rebuild of all packages in unstable, your package failed to build:
>
> NTHREADS = 1
> Setting ITERS_BETWEEN_CHECKPOINTS = 10000.
> INFO: Maximum recommended exponent for FFT length (3840 Kdbl) = 73244853; p[
> = 72123137]/pmax_rec = 0.9846853949.
> Initial DWT-multipliers chain length = [short] in carry step.
> M72123137: using FFT length 3840K = 3932160 8-byte floats, initial residue
> shift count = 38904293
> This gives an average 18.341862233479816 bits per digit
> Using complex FFT radices 240 16 16 32
> Using 1 threads in carry step
> /bin/bash: line 6: 190871 Segmentation fault ${dir}$tst
> FAIL: tests/self-test
> 1 of 3 tests failed
> About the archive rebuild: The build was made on virtual machines from AWS,
> using sbuild and a reduced chroot with only build-essential packages.
>
> If you could not reproduce the bug please contact me privately, as I
> am willing to provide ssh access to a virtual machine where the bug is
> fully reproducible.
Hello Santiago,
The package build on my laptop, on barriere.debian.org and on the
Debian buildd, so it is fair to say that I cannot reproduce this bug.
Does it FTFBS also on the real hardware ?
> Note: I tried 100 times, on 25 different AWS instances, and it failed 100
> times.
> (These days I'm using instances of type c7a, m7a, and r7a, all of which have
> AMD CPUs). I suspect that this could be due to Intel/AMD differences, but if
> that's the case, it would be actually concerning, as it would mean the package
> could not work properly on AMD machines. I'd like to believe that we are not
> lowering support for AMD just because our autobuilders run Intel.
The package can be built! It is only the test-suite that fail on AMD.
The alternative to downgrading this bug is not running the test-suite or
removing the package from Debian.
We can also try an new upstream version but there are no garrantee it works
any differently.
> Also, note that there is a segfault in one of the tests. This is also
> concerning as it means we are letting a script which segfaults (for unknown
> reasons) to decide about whether the program is ok or not.
The test-suite check whether the program has been miscompiled and if the host
is reliable.
Cheers,
--
Bill. <ballo...@debian.org>
Imagine a large red swirl here.
