[Pkg-rust-maintainers] Bug#1079387: rust-typenum - test failure on i386.

[Peter Green]

Package: rust-typenum
Severity: serious

rust-typenum is suffering from a failing test on i386, this
has been the case for some time as evidenced by debci and
reproducible builds test history but did not come to my
attention until a new version was uploaded and a "fails
to migrate to testing for a long time" bug was filed.

I tried to dig into the bug by making the assert more
verbose, and that is when things started to get weird,
the calculated value *does* seem to be less than the
epsilon (specifically the calculated value appears to
be zero) but the assert neverthless fails. The patch I
used for this is in debcargo-conf if anyone else wants
to play with it.

I presume this was triggered by an update of rustc,
but I'm not sure exactly which one. Looking at the
dates of the tests in testing and unstable and comparing
them to rustc's history on the tracker it looks like the
last successful test probablly used 1.66 and the
first failing test probablly used 1.70.

If someone wants to try and write up a compiler bug
based on this they are welcome to do so, I don't have
the energy to do that right now.

Since this is not a regression with respect to testing
and the root cause is probably not in this package,
if there is no better solution I intend to skip the
test. If/when I do so I will downgrade this bug to
"important".