Control: tags -1 patch pending

Greetings,

the remaining failure affects i386 which, from my experiments,
seems to be caused by the excess double precision.  I consider
pushing an adjustment to the test suite, so it builds with that
option, if there are no objections.

Have a nice day,  :)
-- 
Étienne Mollier <emoll...@emlwks999.eu>
Fingerprint:  8f91 b227 c7d6 f2b1 948c  8236 793c f67e 8f0d 11da
Sent from /dev/pts/7, please excuse my verbosity.

diff --git a/debian/tests/test-script b/debian/tests/test-script
index 0897ac6a..ded1b270 100755
--- a/debian/tests/test-script
+++ b/debian/tests/test-script
@@ -3,8 +3,10 @@ set -e
 set -x
 # Choose location for test executable
 e=$(mktemp --tmpdir=${AUTOPKGTEST_TMP} ensmallen-test-XXXXXX)
+# Some tests are sensitive to the excess in double precision of i386.
+test "$(dpkg --print-architecture)" != "i386" || FLOAT_STORE=-ffloat-store
 # Build outside tests to reduce possibility of getting build rather
 # than installed ensmallen files via #include.
-c++ -O -Wall -o ${e} tests/*.cpp -pthread $(pkg-config armadillo --cflags 
--libs) -lpthread
+c++ -O -Wall $FLOAT_STORE -o ${e} tests/*.cpp -pthread $(pkg-config armadillo 
--cflags --libs) -lpthread
 # cd tests because the executable reads data/* files.
 cd tests && ${e} --durations yes

Attachment: signature.asc
Description: PGP signature

Reply via email to