Christopher Faulet

Indeed, an alert is emitted if the request is aborted. But on my side, the same happens with a `lua-load`. And It should be. About the `CR` is is also...

@Darlelet, There is no backport instruction in the commit. Could we close the issue ?

Thanks Felipe. It is really useful because it seems indeed that commits of the 3.0.5 are not responsible of the Luke's issue. Is the bug arise with H1 only or...

Thanks Luke. I don't know if it is the same issue or not. I tried hard to reproduce the issue by hacking the code without success for now. I try...

Indeed, it is possible. But in that case, it is strange to observe a difference when the haproxy reporting errors is downgraded to 3.0.4. For now, I'm unable to reproduce...

So, it suggest the response was fully received by haproxy but something was dropped during the forwarding to the client. It is most probably the last bytes. So I must...

@felipewd, it may be another issue but it is hard to say. Indeed, Luke's bug seems to be related to the 3.0.5. But there are several haproxy in the chain....

It is indeed specific to the splicing, so it should not solve your issue unfortunately...

Well that said, I don't know how your upstream HAProxy is configured. The one reporting `SD--` responses is not affected by this bug. But it may be just the consequence...

@lseelenbinder many thank for the info ! @felipewd copy that, thanks too ! @vuori, that for the report. Just to be clear, on HAproxy side, you observed some `SD--` but...