This is an automated email from the ASF dual-hosted git repository.
dblevins pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/tomee.git
The following commit(s) were added to refs/heads/main by this push:
new 0b0273009b Philosophy and goals of the itests
0b0273009b is described below
commit 0b0273009b5853fa5aedc52b9604ba853448418a
Author: David Blevins <[email protected]>
AuthorDate: Thu Sep 29 23:07:53 2022 -0500
Philosophy and goals of the itests
---
itests/README.adoc | 23 +++++++++++++++++++++++
1 file changed, 23 insertions(+)
diff --git a/itests/README.adoc b/itests/README.adoc
new file mode 100644
index 0000000000..e1b658bc88
--- /dev/null
+++ b/itests/README.adoc
@@ -0,0 +1,23 @@
+# iTest
+
+The goal of the itests are to test and use the TomEE binaries exactly how
users use them; unpack the server, update configurations, deploy an application
and send it requests. When tests are done, we're asserting responses, files on
disk, log output and all expected behavior of the server.
+
+## Why not use Arquillian?
+
+The focus of Arquillian is to create tests for applications that are
abstracted from the server details. iTests are the opposite -- we're testing
the server itself and cannot do that while also abstracting ourselves from the
server.
+
+It's the job of an iTest to manipulate the server exactly the way users would
and assert that the user experience is what we want it to be. Some examples:
+
+- If we tell users, "throw this file in this directory and it should work",
there is ideally an itest that sets up a server, puts that file in the
directory and asserts it does in fact work.
+- If we tell a user, "turn on debug logging and you should see these details",
there is ideally an itest that tries this and asserts the expected debug
messages do show up.
+- If we tell a user, "set this flag to X and it should enable Y behavior",
ideally there is an itest that not only tests the behavior when the flag is
enabled, but tests the behavior when the flag is not enabled and possibly even
tests the error handling when the flag is set to an invalid value.
+
+We're not testing applications, we're testing the server. We don't want to be
abstracted from the server and we don't any magic plumbing like Arquillian's
adapters, deployment code or local/remote testing code. Arquillian is an
awesome hammer, but this problem is not a nail. It solves a different problem.
+
+## Why no dependencies on internals?
+
+While we do not want to be abstracted from the server, we still want to limit
ourselves to using the server the way users would use it. We should not touch
any internal API that is forbidden or off-limits to users. iTests should only
use server features that are supported and considered user-facing. Should
users be using the code in openejb-core? No, they should not, so neither should
itests. Remember, when you're writing an itest the code inside the server is
not "our" code -- you n [...]
+
+Not only is this good for ensuring the user experience always works as
expected, it also means we can theoretically throw alway all the existing TomEE
code, rewrite it entirely and verify user's applications and configuration
file's still work. There are advantages to testing strictly from the user's
perspective and not digging into guts like unit tests do. If we wanted to do
something large and aggressive like cut the size of the server in half, the
itests should continue to work with [...]
+
+A good itest should work even if the server is entirely rewritten.
