This seems to have the wrong build attached to it: 3.4.3.

Thanks, clicked on the wrong one. Fixed.

Works

Works great! LGTM! =)

If you retrigger the tests they would pass. But do beware that the tests seem very minimal as they did not catch the failures in 3.4.3. Is everything really working well now?

There still seem to be regression reports pouring in https://github.com/RsyncProject/rsync/issues/1002

It works for me, but I only tried it against the identical version on the other end.

no regressions noted

@lecris - it is quite impossible to test for every possible use case, even if I wanted to. We have much more tests in RHEL where I tested the 3.4.4 version recently and they all pass. I am monitoring the Upstream github and I'll make more updates if there are more regressions. All in all I think that this update is alright, compared to the 3.4.3 that I unpushed last time.

The AVC failure looks relevant

----
type=AVC msg=audit(06/23/26 07:24:23.201:1266) : avc:  denied  { setpgid } for  pid=7863 comm=timeout scontext=system_u:system_r:sshd_session_t:s0-s0:c0.c1023 tcontext=system_u:system_r:sshd_session_t:s0-s0:c0.c1023 tclass=process permissive=1 
----
type=AVC msg=audit(06/23/26 07:24:23.209:1267) : avc:  denied  { name_connect } for  pid=7865 comm=curl dest=80 scontext=system_u:system_r:sshd_session_t:s0-s0:c0.c1023 tcontext=system_u:object_r:http_port_t:s0 tclass=tcp_socket permissive=1

Are you tracking it?

Also the rpminspect results are worth forwarding to upstream

Forbidden function symbols found:
    gethostbyname
    inet_ntoa

Forbidden symbols were found in an ELF file in the package. The configuration settings for rpminspect indicate the named symbols are forbidden in packages. If this is deliberate, you may want to disable the badfuncs inspection. If it is not deliberate, check the man pages for the named symbols to see what API functions have replaced the forbidden symbols. Usually a function is marked as deprecated but still provided in order to allow for backwards compatibility. Whenever possible the deprecated functions should not be used.

I have the rpminspect check turned off in rawhide - https://src.fedoraproject.org/rpms/rsync/blob/rawhide/f/rpminspect.yaml - both usages are deliberate either as a fallback for systems where there is no better alternative or after IPv6 have already been tried.

I don't think the AVCs are related to rsync, feel free to elaborate if you think they are. This is rather the test design flaw (I will look into that). One AVC is from curl the other is most likely from the beaker itself, the setting of a timeout might trigger that but like I said, I will take a good look at the test.

I have the rpminspect check turned off in rawhide - https://src.fedoraproject.org/rpms/rsync/blob/rawhide/f/rpminspect.yaml - both usages are deliberate either as a fallback for systems where there is no better alternative or after IPv6 have already been tried.

Well you did not port it to the branches

I don't think the AVCs are related to rsync, feel free to elaborate if you think they are. This is rather the test design flaw (I will look into that). One AVC is from curl the other is most likely from the beaker itself, the setting of a timeout might trigger that but like I said, I will take a good look at the test.

The thing that raises my suspicion is that I do not see a similar issue avc issue across other tests. Granted, many of them did not enable avc check. The failure looks very ssh related, which is why it looks suspicious. The root cause might not be here, but knowing which component caused this failure would help point on the update that caused this

Working well, incl SSH transfers.