aboutsummaryrefslogtreecommitdiff
path: root/doc
diff options
context:
space:
mode:
Diffstat (limited to 'doc')
-rw-r--r--doc/develop/tests_sandbox.rst69
-rw-r--r--doc/usage/cmd/ut.rst11
2 files changed, 79 insertions, 1 deletions
diff --git a/doc/develop/tests_sandbox.rst b/doc/develop/tests_sandbox.rst
index 8e42a32afb..bfd3bdb927 100644
--- a/doc/develop/tests_sandbox.rst
+++ b/doc/develop/tests_sandbox.rst
@@ -143,6 +143,75 @@ For example::
Test dm_test_rtc_reset failed 3 times
+Isolating a test that breaks another
+------------------------------------
+
+When running unit tests, some may have side effects which cause a subsequent
+test to break. This can sometimes be seen when using 'ut dm' or similar.
+
+You can use the `-I` argument to the `ut` command to isolate this problem.
+First use `ut info` to see how many tests there are, then use a binary search to
+home in on the problem. Note that you might need to restart U-Boot after each
+iteration, so the `-c` argument to U-Boot is useful.
+
+For example, let's stay that dm_test_host() is failing::
+
+ => ut dm
+ ...
+ Test: dm_test_get_stats: core.c
+ Test: dm_test_get_stats: core.c (flat tree)
+ Test: dm_test_host: host.c
+ test/dm/host.c:71, dm_test_host(): 0 == ut_check_delta(mem_start): Expected 0x0 (0), got 0xffffcbb0 (-13392)
+ Test: dm_test_host: host.c (flat tree)
+ Test <NULL> failed 1 times
+ Test: dm_test_host_dup: host.c
+ Test: dm_test_host_dup: host.c (flat tree)
+ ...
+
+You can then tell U-Boot to run the failing test at different points in the
+sequence:
+
+ => ut info
+ Test suites: 21
+ Total tests: 645
+
+::
+
+ $ ./u-boot -T -c "ut dm -I300:dm_test_host"
+ ...
+ Test: dm_test_pinctrl_single: pinmux.c (flat tree)
+ Test: dm_test_host: host.c
+ test/dm/host.c:71, dm_test_host(): 0 == ut_check_delta(mem_start): Expected 0x0 (0), got 0xfffffdb0 (-592)
+ Test: dm_test_host: host.c (flat tree)
+ Test dm_test_host failed 1 times (position 300)
+ Failures: 4
+
+So it happened before position 300. Trying 150 shows it failing, so we try 75::
+
+ $ ./u-boot -T -c "ut dm -I75:dm_test_host"
+ ...
+ Test: dm_test_autoprobe: core.c
+ Test: dm_test_autoprobe: core.c (flat tree)
+ Test: dm_test_host: host.c
+ Test: dm_test_host: host.c (flat tree)
+ Failures: 0
+
+That succeeds, so we try 120, etc. until eventually we can figure out that the
+problem first happens at position 82.
+
+ $ ./u-boot -T -c "ut dm -I82:dm_test_host"
+ ...
+ Test: dm_test_blk_flags: blk.c
+ Test: dm_test_blk_flags: blk.c (flat tree)
+ Test: dm_test_host: host.c
+ test/dm/host.c:71, dm_test_host(): 0 == ut_check_delta(mem_start): Expected 0x0 (0), got 0xffffc960 (-13984)
+ Test: dm_test_host: host.c (flat tree)
+ Test dm_test_host failed 1 times (position 82)
+ Failures: 1
+
+From this we can deduce that `dm_test_blk_flags()` causes the problem with
+`dm_test_host()`.
+
Running sandbox_spl tests directly
----------------------------------
diff --git a/doc/usage/cmd/ut.rst b/doc/usage/cmd/ut.rst
index 11c64a1779..a3039634f2 100644
--- a/doc/usage/cmd/ut.rst
+++ b/doc/usage/cmd/ut.rst
@@ -8,10 +8,12 @@ Synopis
::
- ut [-r<runs>] [-f] [<suite> [<test>]]
+ ut [-r<runs>] [-f] [-I<n>:<one_test>] [<suite> [<test>]]
<runs> Number of times to run each test
-f Force 'manual' tests to run as well
+ <n> Run <one test> after <n> other tests have run
+ <one_test> Name of the 'one' test to run
<suite> Test suite to run, or `all`
<test> Name of single test to run
@@ -35,6 +37,13 @@ Manual tests are normally skipped by this command. Use `-f` to run them. See
See :ref:`develop/tests_writing:mixing python and c` for more information on
manual test.
+When running unit tests, some may have side effects which cause a subsequent
+test to break. This can sometimes be seen when using 'ut dm' or similar. To
+fix this, select the 'one' test which breaks. Then tell the 'ut' command to
+run this one test after a certain number of other tests have run. Using a
+binary search method with `-I` you can quickly figure one which test is causing
+the problem.
+
Generally all tests in the suite are run. To run just a single test from the
suite, provide the <test> argument.