Dear maintainer(s),

Your package has an autopkgtest, great. However, it fails on s390x (our
only big endian architecture). Can you please investigate the situation
and fix it? I copied some of the output at the bottom of this report. I
strongly suspect the failure is due to endianess, but I'm not sure if
it's only a test failure, or if the package won't work on big endian
systems.

More information about this bug and the reason for filing it can be
found on
https://wiki.debian.org/ContinuousIntegration/RegressionEmailInformation

Paul

[1] https://lists.debian.org/debian-devel-announce/2019/07/msg00002.html

115s ── Failure ('test_h5writeAttributes.R:96:5'): Adding attribute to
dataset ──────
115s attr_back$int_attr[1] inherits from `'NULL'` not `'character'`.
115s ── Failure ('test_h5writeAttributes.R:97:5'): Adding attribute to
dataset ──────
115s attr_back$numeric_attr[1] inherits from `'NULL'` not `'character'`.
115s ── Failure ('test_h5writeAttributes.R:130:5'): Checking other
string options when adding attributes ──
115s `attr_back` has length 0, not length 4.
115s ── Failure ('test_h5writeAttributes.R:133:5'): Checking other
string options when adding attributes ──
115s sort(expected) not identical to sort(names(attr_back)).
115s Types not compatible: character is not NULL
115s ── Failure ('test_h5writeAttributes.R:134:5'): Checking other
string options when adding attributes ──
115s unname(unlist(attr_back[expected])) not identical to c("blah",
"blah2", "blah3", "blah4").
115s target is NULL, current is character
115s ── Failure ('test_h5writeAttributes.R:149:5'): Overwrite exisiting
attribute ───
115s attr_list$char_attr[1] not identical to "new_character".
115s target is NULL, current is character
115s
115s [ FAIL 38 | WARN 0 | SKIP 2 | PASS 974 ]

Hello all,

Most of bioconductor 3.19 has migrated to testing, but some legitimate bugs shown by running the r-bioc-rhdf5 tests on s390x are holding up 20+ packages.

I have communicated the issue to upstream, providing them with a reproducer, and helping them reproduce it on their CI. https://github.com/grimbough/rhdf5/issues/138

However I don't think it will be fixed soon, and while I have made a new release of r-bioc-rhdf5 that skips those tests, I don't feel good about that.

So I propose to add a "architecture-is-little-endian" build-dep to the following packages and request the removal of their s390x builds from the archive:

r-bioc-rhdf5
r-bioc-alabaster.base
r-bioc-alabaster.matrix
r-bioc-bsseq
r-bioc-dropletutils
r-bioc-hdf5array
r-bioc-dss
r-bioc-glmgampoi
r-bioc-gsva
r-bioc-demixt

Two of these packages also have failing s390x tests: r-bioc-alabaster.matrix and r-bioc-alabaster.base

I think the removal can be done via one request, so the only real pain is making new releases of those 11 packages.

Thoughts, comments?