Check the test-suite specific coverage file, and upload it earlier in the process

pavelib/tests.py

@@ -331,12 +331,14 @@ def coverage():
     report_dir = Env.REPORT_DIR
     rcfile = Env.PYTHON_COVERAGERC
 
-    if not (report_dir / '.coverage').isfile():
+    combined_report_file = report_dir / '{}.coverage'.format(os.environ.get('TEST_SUITE', ''))
+
+    if not combined_report_file.isfile():
         # This may be that the coverage files were generated using -p,
         # try to combine them to the one file that we need.
         sh(u"coverage combine --rcfile={}".format(rcfile))
 
-    if not os.path.getsize(report_dir / '.coverage') > 50:
+    if not os.path.getsize(combined_report_file) > 50:
         # Check if the .coverage data file is larger than the base file,
         # because coverage combine will always at least make the "empty" data
         # file even when there isn't any data to be combined.

scripts/jenkins-report.sh

 #!/usr/bin/env bash
 set -e
+set -x
 
 # This script generates coverage and diff cover reports, and optionally
 # reports this data to codecov.io. The following environment variables must be
@@ -14,6 +15,10 @@ source scripts/jenkins-common.sh
 # Get the diff coverage and html reports for unit tests
 paver coverage -b $TARGET_BRANCH
 
+if [[ -n "$PYTEST_CONTEXTS" ]]; then
+    paver upload_coverage_to_s3
+fi
+
 # Test for the CodeCov API token
 if [ -z $CODE_COV_TOKEN ]; then
     echo "codecov.io API token not set."
@@ -27,7 +32,3 @@ fi
 # JUnit test reporter will fail the build
 # if it thinks test results are old
 touch `find . -name *.xml` || true
-
-if [[ -n "$PYTEST_CONTEXTS" ]]; then
-    paver upload_coverage_to_s3
-fi