Signed-off-by: Flavia Rainone <[email protected]>

Signed-off-by: Flavia Rainone <[email protected]>

…act and download-artifact actions to v4

Signed-off-by: Flavia Rainone <[email protected]>

…nd adding a response listener only after push resource has been handled by the listener

This will prevent the race condition where shutdown is processed by the server before all the other steps, which would prevent those steps from occurring

Signed-off-by: Flavia Rainone <[email protected]>

… when investigating PushResourceRSTTestCase

Signed-off-by: Flavia Rainone <[email protected]>

…failures in CI.

Signed-off-by: Flavia Rainone <[email protected]>

…tion after already loading parts

…nto the status line.

Add tests for this and the default behavior.

Signed-off-by: Flavia Rainone <[email protected]>

…rror if buffer.remaining() == remaining after flush. We need to update remaining to do a proper check here.

The aforementioned == check can return true after a successfull flush, because "remaining" value is set to be the size of the bytes remaining in the buffer before encoding. If, at that stage, the buffer is empty, "remaining" value is set to buffer.capacity(). When encoding many bytes, we will end up with a full buffer. As a result, we will try to flush, and it is often the case that the buffer will be entirely cleared after flushing. This results in buffer.remaining() being the overall buffer size, i.e., buffer.capacity(). As a result, buffer.remaining() == remaining check returns true, and write is aborted, causing the broken responses we are seeing.

Signed-off-by: Flavia Rainone <[email protected]>

[UNDERTOW-2412 / 2445 / 2449 / 2446 / 2436 / 2422 / 2444 / 2448] Backport fixes to 2.3.x

[UNDERTOW-2333_2.3.x] Undertow read/write timeout should not apply to WebSockets or SSE

Signed-off-by: Flavia Rainone <[email protected]>