Bug#1135798: openssh-client: OpenSSH in trixie-p-u breaks ControlPersist
Colin Watson <[email protected]> Wed, 6 May 2026 13:00:33 +0100
| Newsgroups | gmane.linux.debian.devel.ssh |
|---|---|
| Message-ID | <afstYXL3o9NczDcq__32989.560793952$1778069020$gmane$org@riva.ucam.org> |
On Wed, May 06, 2026 at 12:06:03PM +0100, Colin Watson wrote: >On Tue, May 05, 2026 at 09:42:31PM -0400, Jon wrote: >>I'm not sure if this is the right way to file a bug against something >>sitting in stable-proposed-updates. I'm flagging it as important only >>because it would be a notable regression if it reached stable. > >Oops - thanks for the heads-up! CCing #1135624 so that the release >team is aware of the regression. > >>The recent upload of OpenSSH to trixie-p-u backported the IPQoS changes >>from 10.1p1 without including the fix for bz#3872 >> >>https://bugzilla.mindrot.org/show_bug.cgi?id=3872 >> >>https://anongit.mindrot.org/openssh.git/commit/?h=V_10_1&id=979cbc2c1e0c9cd2f60d45d8d1da69519ec425cf >> >>I've confirmed that the bug appears in the package sitting in trixie-p-u > >I'm test-building the attached patch, and will upload to trixie-p-u if >it checks out. This works. Manual test plan: in a fresh container with openssh-client installed, use `ssh -S"$(pwd)/sock" -o ControlMaster=auto -o ControlPersist=5s` to a target host. The version currently in trixie-p-u has horrible delays as described in the upstream bug, which go away with that patch.