base r344601 introduced a new feature to update zfs send process title with the send progress, but at the same time it appears to have broken -n -P combination of option that's used to estimate a send stream size (while doing a "dry" send) and to report it in a machine readable form. On a system without the change: > $ zfs send -n -v -i testz/test@snap1 testz/test@snap2 > send from @snap1 to testz/test@snap2 estimated size is 2.01M > total estimated size is 2.01M > $ zfs send -n -P -i testz/test@snap1 testz/test@snap2 > incremental snap1 testz/test@snap2 2109424 > size 2109424 On a system with the change: > $ zfs send -n -v -i testz/test@snap1 testz/test@snap2 > send from @snap1 to testz/test@snap2 estimated size is 2.01M > total estimated size is 2.01M > $ zfs send -n -P -i testz/test@snap1 testz/test@snap2 > incremental snap1 testz/test@snap2 As can be seen, "-n -v" works as before, but "-n -P" is broken.
This is a potential blocker for 12.1 release as this is a regression since 12.0.
Proposed patch in https://reviews.freebsd.org/D21709. If I don't get more feedback from it by tomorrow, I'll commit it over the weekend, and MFC 2 weeks after that, unless things need to move faster.
For any issues added as a bug 240700 blocker, please either include the patch as an attachment with maintainer-approval flag set to ? re@, or add re@ as a blocking reviewer on the differential
For the review, re@ doesn't work; releng?
Created attachment 207644 [details] Proposed patch Proposed diff, as per request.
(In reply to Kubilay Kocak from comment #3) As mentioned in (many) emails to developers, re@ does not currently accept Phabricator reviews for approval requests. Please see https://wiki.freebsd.org/Releng/ChangeRequestGuidelines for the procedures.
(In reply to Glen Barber from comment #6) Well, this is not a merge request (yet). The fix is not even in head. I am sure that when the time comes to merge from stable/12 to releng/12.1, there will be a proper approval request. I am not sure why there are some many requirements and requests to sef at *this* stage.
(In reply to Andriy Gapon from comment #7) I misunderstood the context of the PR. I apologize for the confusion.
A commit references this bug: Author: sef Date: Sat Sep 21 17:54:42 UTC 2019 New revision: 352580 URL: https://svnweb.freebsd.org/changeset/base/352580 Log: Fix a regression introduced in r344601, and work properly with the -v and -n options. PR: 240640 Reported by: Andriy Gapon <avg@FreeBSD.org> Reviewed by: avg MFC after: 2 weeks Differential Revision: https://reviews.freebsd.org/D21709 Changes: head/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_sendrecv.c
Sean, is it possible to MFC this change to stable/12 a little bit sooner, so that we can try to merge it to releng/12.1 while it's still in the RC phase? Thanks!
(In reply to Andriy Gapon from comment #10) Sorry, I meant beta phase, of course.
I completely forgot about it this weekend. I'll plan on doing things today.
A commit references this bug: Author: sef Date: Tue Oct 1 18:05:52 UTC 2019 New revision: 352932 URL: https://svnweb.freebsd.org/changeset/base/352932 Log: Fix a regression introduced in r344601, and work properly with the -v and -n options. PR: 240640 Reported by: Andriy Gapon <avg@FreeBSD.org> Reviewed by: avg Differential Revision: https://reviews.freebsd.org/D21709 Approved by: re Changes: releng/12.1/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_sendrecv.c
Can this PR be closed?
A commit references this bug: Author: avg Date: Wed Oct 9 09:47:00 UTC 2019 New revision: 353338 URL: https://svnweb.freebsd.org/changeset/base/353338 Log: MFC r352580 by sef: Fix a regression introduced in r344601 ... and work properly with the -v and -n options. PR: 240640 Changes: _U stable/11/ stable/11/cddl/contrib/opensolaris/lib/libzfs/common/libzfs_sendrecv.c
(In reply to Mark Johnston from comment #14) Yes, of course. Let me do it now.