openldap: update to 2.6.14 - #30357
Conversation
|
Looks good. Recommend merge. |
openwrt-ai
left a comment
openwrt-ai
left a comment
There was a problem hiding this comment.
Reviewed 1 new commit. The version/hash change itself looks correct (PKG_RELEASE is already 1, and CI's download step validated the new PKG_HASH), but the build fails on every arch because two patches are dirty against 2.6.14 — details inline.
Generated by Claude Code
|
|
||
| PKG_NAME:=openldap | ||
| PKG_VERSION:=2.6.13 | ||
| PKG_VERSION:=2.6.14 |
There was a problem hiding this comment.
Two of the in-tree patches no longer apply cleanly to the 2.6.14 tarball, and all 10 Feeds Package Test Build jobs fail on it (aarch64_generic, x86_64, mips_24kc, mipsel_24kc, arm_cortex-a9, arm_cortex-a15, riscv64_generic, i386_pentium-mmx, powerpc_464fp, powerpc_8548 — same failure everywhere):
Applying patch 001-automake-compat.patch
patching file servers/slapd/Makefile.in
Hunk #1 succeeded at 13 with fuzz 1.
Applying patch 750-no-strip.patch
patching file servers/slapd/Makefile.in
Hunk #1 succeeded at 366 (offset -11 lines).
...
Dirty patches detected, please refresh and review the diff
The refresh step regenerated 001-automake-compat.patch and 750-no-strip.patch, so those two need to be refreshed against 2.6.14 (make package/openldap/refresh) and the resulting diff committed alongside the version bump. The other four patches came back unchanged.
For the record, the download step did fetch openldap-2.6.14.tgz and validate it against the new PKG_HASH, so the hash update itself is fine.
Generated by Claude Code
454e35b to
d979c0b
Compare
d979c0b to
dfd4731
Compare
openwrt-ai
left a comment
openwrt-ai
left a comment
There was a problem hiding this comment.
Reviewed 1 new commit (force-push 454e35b → dfd4731). The only change since the last review is a git format-patch-style header added to 001-automake-compat.patch and 750-no-strip.patch; the version/hash bump is untouched and still looks right (PKG_RELEASE is 1, correct for a version bump). Two points inline: the patch bodies were not refreshed, so the 2.6.14 dirty-patch build failure is unchanged, and the new headers put a different author on two pre-existing patches. Neither is a correctness problem in the packaged software — the refresh is the one that blocks a build.
Generated by Claude Code
| From: Romain Isnel <[email protected]> | ||
| Date: Sun, 23 Aug 2026 12:23:37 +0200 | ||
| Subject: [PATCH] automake compat | ||
|
|
||
| Signed-off-by: Romain Isnel <[email protected]> |
There was a problem hiding this comment.
These headers attribute an existing patch to a new author. 001-automake-compat.patch is not new in this PR — it arrived with 3c78eaf "Copy openldap package from old repository" (2014), was extended in ba903d4 and 6d921a8, and last touched in d28c3a2 — all by W. Michael Petullo / Val Kulkov, none by this PR. From: is what git am records as the patch author and Signed-off-by: certifies the origin of the change, so adding both with a different name rewrites the authorship of someone else's patch.
Since both of these are OpenWrt-local build hacks with no upstream to be sent to, the guideline that asks for a full header ("structured in a way that it is potentially upstreamable") does not really apply to them — dropping the From: and Signed-off-by: lines and keeping only Subject:/Date:, or setting From: to the original author, would avoid the misattribution.
Same applies to lines 2 and 6 of 750-no-strip.patch, which originates from the same 2014 import.
Generated by Claude Code
| Subject: [PATCH] no strip | ||
|
|
||
| Signed-off-by: Romain Isnel <[email protected]> | ||
| --- |
There was a problem hiding this comment.
The new commit adds the header block but leaves the patch body byte-identical (git diff 454e35b..dfd4731 on this file is the 7 added header lines and nothing else), so the fuzz/offset problem reported earlier in this PR is still there: against 2.6.14 the servers/slapd/Makefile.in hunk of this patch applies at 366 rather than the @@ -377,7 +377,7 @@ recorded below, and 001-automake-compat.patch applies to the same file with fuzz 1. Both need to be regenerated against the new tarball with make package/openldap/refresh and the resulting body changes committed, otherwise the build stops at "Dirty patches detected, please refresh and review the diff". Worth doing before the reviewers' build jobs are approved, since the header-only change does not affect that step.
Generated by Claude Code
dfd4731 to
102fa4e
Compare
openwrt-ai
left a comment
openwrt-ai
left a comment
There was a problem hiding this comment.
Reviewed 1 new commit (force-push dfd4731 → 102fa4e). The version/hash bump is untouched and still looks right.
The two patch bodies were refreshed this time: 750-no-strip.patch now records @@ -366,7 +366,7 @@ for servers/slapd/Makefile.in instead of @@ -377,..., and 001-automake-compat.patch dropped .backend from the XPROGRAMS= context line — exactly the two things the earlier "Dirty patches detected" failure pointed at. I could not confirm it against the 2.6.14 tarball from here (no egress beyond GitHub) and the Feeds Package Test Build jobs have not been approved on this head, so only the reviewers' build run can settle it, but the shape of the change is what a make package/openldap/refresh produces.
One real defect inline: 750-no-strip.patch now carries Subject: [PATCH] automake compat, copied from the other patch. Nothing blocking a merge beyond that one-line fix; the Date: remark is optional.
Generated by Claude Code
| From 3c78eaff96d7ba286bb0404b9bdc1a2c6aca7ec0 Mon Sep 17 00:00:00 2001 | ||
| From: W. Michael Petullo <[email protected]> | ||
| Date: Sun, 6 Aug 2014 11:28:00 +0200 | ||
| Subject: [PATCH] automake compat |
There was a problem hiding this comment.
The subject describes the wrong patch: this file drops $(STRIP_OPTS) from the two $(LTINSTALL) calls, it has nothing to do with automake. It looks like the whole header block was copied over from 001-automake-compat.patch — the previous revision of this PR had Subject: [PATCH] no strip here, and the force-push replaced it along with the From:/Date: lines.
| Subject: [PATCH] automake compat | |
| Subject: [PATCH] no strip |
Generated by Claude Code
| @@ -1,3 +1,10 @@ | |||
| From 3c78eaff96d7ba286bb0404b9bdc1a2c6aca7ec0 Mon Sep 17 00:00:00 2001 | |||
| From: W. Michael Petullo <[email protected]> | |||
| Date: Sun, 6 Aug 2014 11:28:00 +0200 | |||
There was a problem hiding this comment.
nit: the Date: does not match the commit named on line 1. 3c78eaf "Copy openldap package from old repository" is authored Fri Jul 18 15:05:59 2014 -0400, and 6 Aug 2014 was a Wednesday, not a Sunday — so the line reads as invented rather than taken from the commit it cites. Either copy the real author date (Fri, 18 Jul 2014 15:05:59 -0400) or drop the Date: line; quilt ignores it either way.
Same line 3 in 750-no-strip.patch, which carries the identical header.
Thanks for switching the attribution back to the original author — that resolves the earlier concern about the From:/Signed-off-by: lines.
Generated by Claude Code
Changelog is available at: https://openldap.org/software/release/changes_lts.html Reworked some patches with headers to comply with patches guidelines. Signed-off-by: Romain Isnel <[email protected]>
102fa4e to
3ab3657
Compare
openwrt-ai
left a comment
openwrt-ai
left a comment
There was a problem hiding this comment.
Reviewed 1 new commit (force-push 102fa4e → 3ab3657); no new issues found.
The only change is the Subject: line of 750-no-strip.patch, automake compat → no strip — the fix asked for in the previous review. The version/hash bump and both refreshed patch bodies are untouched, and all three FormalityCheck jobs are green on this head.
Two earlier threads are still open but neither is new to this revision and neither blocks a merge from where I sit:
- The optional
Date:nit on line 3 of both patches —Sun, 6 Aug 2014doesn't match the commit cited on line 1. Purely cosmetic; quilt ignores the line. - The patch refresh still has not been exercised by a build.
Feeds Package Test Buildhas not been approved on102fa4eor on this head, so the refreshed hunk offsets remain unverified from my side (no egress beyond GitHub here to check them against the 2.6.14 tarball). Approving the build jobs is the thing that would settle it.
Generated by Claude Code
3 participants
📦 Package Details
Maintainer: @flyn-org
Description: OpenLDAP Software is an open source implementation of the Lightweight Directory Access Protocol.
Changelog of the changes in this commit are available at https://openldap.org/software/release/changes_lts.html
Reworked some patches with headers to comply with patches guidelines.
🧪 Run Testing Details
✅ Formalities
If your PR contains a patch:
git am