summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorMaoni0 <maonis@microsoft.com>2016-06-21 15:36:05 -0700
committerMaoni0 <maonis@microsoft.com>2016-06-21 15:36:05 -0700
commitce86b743aefa7decd5490a9f8a86b95dbe23a9d9 (patch)
treead214d95e798335414a65b99f66179ecf188b22f
parent751dcd675ab685386f981d8ec1a2df6b004c19a2 (diff)
downloadcoreclr-ce86b743aefa7decd5490a9f8a86b95dbe23a9d9.tar.gz
coreclr-ce86b743aefa7decd5490a9f8a86b95dbe23a9d9.tar.bz2
coreclr-ce86b743aefa7decd5490a9f8a86b95dbe23a9d9.zip
added a note about the reason why we ignore failures from tests in the stress run infrastructure
-rw-r--r--tests/src/GC/Stress/stress_run_readme.txt2
1 files changed, 2 insertions, 0 deletions
diff --git a/tests/src/GC/Stress/stress_run_readme.txt b/tests/src/GC/Stress/stress_run_readme.txt
index 56b9ff556f..7a2cf2b1ab 100644
--- a/tests/src/GC/Stress/stress_run_readme.txt
+++ b/tests/src/GC/Stress/stress_run_readme.txt
@@ -6,6 +6,8 @@ Most of the code in gc.cpp is intricate - unless you are touching something that
This was adapted from one of the ways we run stress internally. The idea is that it takes a config that specifies the tests to run and runs them all in one process to stress random combinations of allocation and survival patterns.
+Note that these tests were picked from functional tests so some of them could have failures because some condition it checks for is not met. In stress runs, we only care about AVs so we want to run them as long as possible and don't care about failures indicated by the tests themselves.
+
This is a pretty crude implementation. Feel free to improve it!
2. Setting up stress