Ludovic Orban
Ludovic Orban
This was partially fixed in https://github.com/eclipse/jetty.project/commit/cdc1b079ae509e3485200728fd3d7acdce6561b2#diff-72000f31f0a66b20471a919375f90b2c4184d6d05281084fe97f20cdc2505e25R166 We're going to need https://github.com/eclipse/jetty.project/pull/8409 to properly fix this. I'll take care of this once that other PR gets merged.
No, there is nothing I can see that looks suspicious. A `HttpClient` dump would certainly help figuring out what's going on, but sharing your heap dump (if you can) could...
The latest version of `h2spec`'s 8.1.2.6 test accepts `RST_STREAM` as a valid response: https://github.com/summerwind/h2spec/blob/master/http2/8_1_2_6_malformed_requests_and_responses.go#L69-L97 but according to the above error message, the test fails because `GOAWAY` was expected but `RST_STREAM`...
Running the latest h2spec v2.6.0 binary, 8.1.2.6 sometimes fails because it responds to the invalid request with a `DATA` frame instead of a `RST_STREAM` one, see the attached [8_1_2_6_failure.pcap.gz](https://github.com/eclipse/jetty.project/files/8999408/8_1_2_6_failure.pcap.gz) file...
Saw this failure in jenkins: ``` 8.1.2.6. Malformed Requests and Responses using source address 172.17.0.8:40564 × 1: Sends a HEADERS frame with the "content-length" header field which does not equal...
This isn't as easy as it sounds. Editing out `/META-INF/service/...` files on the fly would require handing out `URL`s that point to the modified files. Should we hand out `URL`s...
My personal opinion is that neither `resolve()` nor `getPath()` should ever return null, but I'm open to different opinions.
@gregw sorry, I've had to revert these changes in https://github.com/eclipse/jetty.project/pull/8452 because they broke the build.
The [12.0.x EventsHandler](https://github.com/eclipse/jetty.project/blob/jetty-12.0.x/jetty-core/jetty-server/src/main/java/org/eclipse/jetty/server/handler/EventsHandler.java) has a `void onResponseTrailersComplete(Request request, HttpFields trailers)` event that should cover this.
I think @gregw made a few good points: calling `addBean` from constructors has been the norm forever; we use that pattern extensively, and users may do too. Adding a new...