On 10/23/2022 10:51 PM, Ryan Schmitt wrote:
If you'd like, I could run `git-bisect` to figure out where the tests
broke; I'll just need to know which tests broke and how to run them.
Much appreciated, Ryan. The tests tend to fail intermittently and there
is no guarantee one can reliably catch the regressions during the
git-bisect run. Anyway, I was able to reproduce one of the regressions
with debug logging on [1] and I should be able to fix it. The other one
was likely due to a flaw in data capacity counting in a test message
handler.
Oleg
[1]
https://github.com/apache/httpcomponents-core/actions/runs/3307510738/jobs/5459198319
On Sat, Oct 22, 2022 at 8:14 AM Oleg Kalnichevski <[email protected]> wrote:
The JUnit 5 upgrade was a <self censored> cluster<self censored>.
The original contributor ported some of the easy test cases that did
not have a complex resource setup and left most complex and most
important protocol and integration tests still partially using JUnit 4
features through the JUnit 5 migration layer. I cannot blaim him. JUnit
5 after so many years still have no reasonable way of parameterizing
test cases through constructors. What the original contributor did not
do (by mistake or by oversight) is activating the vintage JUnit support
for Maven Surefire plugin.
As a result we have been running CI of the 5.2 code _without_
integration tests for a _whole_ <self censored> year (since Nov 2021).
In the meantime two integration tests regressed and I have no idea at
what point it has happended and what might have broken them.
It is too late to go back to JUnit 4. I have migrated the remaining
test cases still using the migration support layer. Those test cases
got more verbose and uglier but at least they work as intended with
JUnit 5.
This is a prime example that of one uses newer, supposedly better
frameworks, without any particular practical reason, one ends up with
things not being better, but just different.
Oleg
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
