[R-pkg-team] Bug#1028111: r-cran-clock: autopkgtest failure on 32bit

[Dirk Eddelbuettel]

The 'bug fix' applied to the _previous_ version of CRAN package clock, namely
a somewhat 'wild' sed patching of a single ie

# Ignore single test for i386 architecture.  This is a workaround for bug #1024828 of r-cran-igraph
if [ "$hostarch" = "i386" -o "$hostarch" = "armel" -o "$hostarch" = "armhf" ] ; then
  sed -i -e '152d' -e '151d' testthat/test-posixt.R
fi

now creates a syntax error on these very architectures (as the upstream file
presumably changed).  Here is the corresponding part of one of the logs
(armel at https://ci.debian.net/data/autopkgtest/testing/armel/r/r-cran-clock/36601943/log.gz) 

 80s > library(testthat)
 80s > library(clock)
 80s > 
 80s > test_check("clock")
132s Error in parse(con, n = -1, srcfile = srcfile, encoding = "UTF-8") : 
132s   test-posixt.R:1372:0: unexpected end of input
132s 1370:   })
132s 1371: })
132s      ^
132s Calls: test_check ... doTryCatch -> lapply -> FUN -> source_file -> parse
132s Execution halted
133s autopkgtest [08:19:23]: test run-unit-test: -----------------------]
133s autopkgtest [08:19:23]: test run-unit-test:  - - - - - - - - - - results - - - - - - - - - -
133s run-unit-test        FAIL non-zero exit status 1

Obviously if we ambush upstream code and corrupt then tests we impose will
fail.

This has fairly large repurcussions across other packages include some of mine.

If nobody else steps up I plan to address this by no longer 'hot-fix
patching' (and thereby corrupting) test-posixt.R but (at least for now)
simply skipping tests of test-posixt.t.  The regular maintainers may find
time to investigate what parts of the test file work for armel, armhf, i386.

Dirk





-- 
dirk.eddelbuettel.com | @eddelbuettel | edd at debian.org