aboutsummaryrefslogtreecommitdiffstats
path: root/meta/classes/imagetest-qemu.bbclass
diff options
context:
space:
mode:
authorJiajun Xu <jiajun.xu@intel.com>2010-07-08 02:06:58 +0800
committerRichard Purdie <rpurdie@linux.intel.com>2010-07-15 12:48:25 +0100
commitfd1d661b391dfad1edaac937a17c1165f635031a (patch)
tree3bce526d67faf52385482f38c31679db05f6dbd5 /meta/classes/imagetest-qemu.bbclass
parentbd19169a60d5aaadc278439578d06c8027cfecf3 (diff)
downloadopenembedded-core-fd1d661b391dfad1edaac937a17c1165f635031a.tar.gz
test: add automation framework and sanitytest
Automation test is disabled by default. User need set TESTCLASS to qemu in conf/local.conf and run bitbake command "bitbake poky-image-xxx" or "bitbake poky-image-xxx -c qemuimagetest" to trigger it. Currently only the sanity test with two testcases are added. To run the test, user need prepare a testing environment: 1) "expect" should be installed on system 2) NOPASSWD should be set for user to run bitbake Signed-off-by Jiajun Xu <jiajun.xu@intel.com>
Diffstat (limited to 'meta/classes/imagetest-qemu.bbclass')
-rw-r--r--meta/classes/imagetest-qemu.bbclass137
1 files changed, 137 insertions, 0 deletions
diff --git a/meta/classes/imagetest-qemu.bbclass b/meta/classes/imagetest-qemu.bbclass
new file mode 100644
index 0000000000..742b759208
--- /dev/null
+++ b/meta/classes/imagetest-qemu.bbclass
@@ -0,0 +1,137 @@
+addtask qemuimagetest before do_build
+# after do_rootfs
+do_qemuimagetest[nostamp] = "1"
+do_qemuimagetest[depends] += "qemu-native:do_populate_sysroot"
+
+# Test related variables
+# By default, TEST_DIR is created under WORKDIR
+TEST_DIR ?= "${WORKDIR}/qemuimagetest"
+TEST_LOG ?= "${LOG_DIR}/qemuimagetests"
+TEST_RESULT ?= "${TEST_DIR}/result"
+TEST_LIST ?= "${TEST_DIR}/list"
+TEST_TMP ?= "${TEST_DIR}/tmp"
+TEST_SCEN ?= "sanity"
+
+python do_qemuimagetest() {
+ import sys
+ import re
+ import os
+
+ """
+ Test Controller for Poky Testing.
+ """
+
+ casestr = re.compile(r'(?P<scen>\w+\b)\s*(?P<case>\w+$)')
+ resultstr = re.compile(r'\s*(?P<case>\w+)\s*(?P<pass>\d+)\s*(?P<fail>\d+)\s*(?P<noresult>\d+)')
+
+ """funtion to run each case under scenario"""
+ def runtest(scen, case, fulltestpath):
+ resultpath = bb.data.getVar('TEST_RESULT', d, 1)
+ machine = bb.data.getVar('MACHINE', d, 1)
+ pname = bb.data.getVar('PN', d, 1)
+
+ testpath = bb.data.getVar('TEST_DIR', d, 1)
+
+ """initialize log file for testcase"""
+ logpath = bb.data.getVar('TEST_LOG', d, 1)
+ bb.utils.mkdirhier("%s/%s" % (logpath, scen))
+ caselog = os.path.join(logpath, "%s/log_%s.%s" % (scen, case, bb.data.getVar('DATETIME', d, 1)))
+ os.system("touch %s" % caselog)
+
+
+ """export TEST_TMP, TEST_RESULT, DEPLOY_DIR and QEMUARCH"""
+ os.environ["PATH"] = bb.data.getVar("PATH", d, True)
+ os.environ["TEST_TMP"] = testpath
+ os.environ["TEST_RESULT"] = resultpath
+ os.environ["DEPLOY_DIR"] = bb.data.getVar("DEPLOY_DIR", d, True)
+ os.environ["QEMUARCH"] = machine
+ os.environ["QEMUTARGET"] = pname
+ os.environ["DISPLAY"] = bb.data.getVar("DISPLAY", d, True)
+
+ """run Test Case"""
+ bb.note("Run %s test in scenario %s" % (case, scen))
+ os.system("%s | tee -a %s" % (fulltestpath, caselog))
+
+ """Generate testcase list in runtime"""
+ def generate_list(testfile, testlist):
+ list = []
+ if len(testlist) == 0:
+ raise bb.build.FuncFailed("No testcase defined in TEST_SCEN")
+
+ """remove old testcase list file"""
+ if os.path.exists(testfile):
+ os.remove(testfile)
+ os.system("touch %s" % testfile)
+
+ """check testcase folder and add case to TEST_LIST"""
+ for item in testlist.split(" "):
+ found = False
+ for dir in bb.data.getVar("QEMUIMAGETESTS", d, True).split():
+ casepath = os.path.join(dir, item)
+ if not os.path.isdir(casepath):
+ continue
+ files = os.listdir(casepath)
+ for casefile in files:
+ fulltestcase = "%s/%s" % (casepath, casefile)
+ if os.path.isfile(fulltestcase):
+ list.append((item, casefile, fulltestcase))
+ found = True
+ if not found:
+ raise bb.build.FuncFailed("Testcase folder not found for test %s" % item)
+ return list
+
+ """check testcase folder and create test log folder"""
+ testpath = bb.data.getVar('TEST_DIR', d, 1)
+ bb.utils.mkdirhier(testpath)
+
+ logpath = bb.data.getVar('TEST_LOG', d, 1)
+ bb.utils.mkdirhier(logpath)
+
+ tmppath = bb.data.getVar('TEST_TMP', d, 1)
+ bb.utils.mkdirhier(tmppath)
+
+ """initialize result file"""
+ resultpath = bb.data.getVar('TEST_RESULT', d, 1)
+ bb.utils.mkdirhier(resultpath)
+ resultfile = os.path.join(resultpath, "testresult.%s" % bb.data.getVar('DATETIME', d, 1))
+ sresultfile = os.path.join(resultpath, "testresult.log")
+
+ machine = bb.data.getVar('MACHINE', d, 1)
+
+ if os.path.exists(sresultfile):
+ os.remove(sresultfile)
+ os.system("touch %s" % resultfile)
+ os.symlink(resultfile, sresultfile)
+ f = open(sresultfile, "a")
+ f.write("\tTest Result for %s\n" % machine)
+ f.write("\tTestcase\tPASS\tFAIL\tNORESULT\n")
+ f.close()
+
+ """generate pre-defined testcase list"""
+ testfile = bb.data.getVar('TEST_LIST', d, 1)
+ testlist = bb.data.getVar('TEST_SCEN', d, 1)
+ fulllist = generate_list(testfile, testlist)
+
+ """Begin testing"""
+ for test in fulllist:
+ (scen, case, fullpath) = test
+ runtest(scen, case, fullpath)
+
+ """Print Test Result"""
+ ret = 0
+ f = open(sresultfile, "r")
+ for line in f:
+ m = resultstr.match(line)
+ if m:
+ if m.group('fail') == "1":
+ ret = 1
+ elif m.group('noresult') == "1":
+ ret = 2
+ print line,
+ else:
+ print line,
+ f.close()
+
+ if ret != 0:
+ raise bb.build.FuncFailed("Some testcases fail, pls. check test result and test log!!!")
+}