[WIP] http4s - migrate from Blaze to Ember#4185
Conversation
16a531e to
cd1ce53
Compare
088bbfa to
6aa2b66
Compare
f0b3c83 to
57c2fa6
Compare
|
Oh nice, I was worried you abandoned the PR ;) |
|
Will wait for #4267 first |
|
@ivan-klass sure thing, done :) |
3fe8b8e to
8b5ea8a
Compare
|
To whom it may concern: ember builder doesn't have an option for a specific execution context, so test request handling is not in the global one as it was before. It seems like race conditions are observed when dealing with some web-socket tests. |
|
@ivan-klass sorry for the delay - the global EC was used before only because it was the easiest one to obtain ;). I don't think there's anything in the tests that would rely on the global one being used: each server should be independent anyway, regardless on the thread pool on which tasks/futures are scheduled. So I would look for causes of the exceptions elsewhere, unfortunately. |
|
Then there might be a bug in http4-ember then... I'll try to dig in and reproduce when I have time |
195128a to
7b7eda3
Compare
…mber Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
The 5s idle-timeout override cancels in-flight chunked-response writes under load, truncating the response before the terminal chunk and making the client see an EOF mid-read (http4s#6427). This surfaced as a flaky streaming test after the blaze->ember migration. Teardown is handled by withShutdownTimeout(0), so the short timeout is no longer needed for test speed. Also add the noDelay socket option to the ZIO interpreter for parity with the cats one (http4s#7668). Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Ember seems to be more modern and especially better supported in terms of cross-platform builds