Heimdal can be synced

[Michał Małoszewski]

After heimdal merge process, I was trying to figure out if the delta that is still not dropped is required or not.
So, to test it, I have created 2 PPAs, one in which lto is disabled, and the second one, where the lto is enabled. I have built them and downloaded the debs for i386 and amd64.
Then I have compared amd64 deb from PPA1 with amd64 deb from PPA2. The same story with i386. The binary dependencies were identical. There is no difference between the files. So in that case, the delta can possibly be dropped.

The package has already been merged again, with the change:

heimdal (7.8.git20221117.28daf24+dfsg-2ubuntu1) mantic; urgency=low

  * Merge from Debian unstable. Remaining changes:
    - d/rules: Disable lto, to regain dep on roken, otherwise
      dependencies on amd64 are different than i386 resulting in
      different files on amd64 and i386.

 -- Steve Langasek <email address hidden> Tue, 02 May 2023 09:56:10 +0200

heimdal (7.8.git20221117.28daf24+dfsg-1ubuntu1) lunar; urgency=low

  * Merge from Debian unstable. Remaining changes:
    - d/rules: Disable lto, to regain dep on roken, otherwise
      dependencies on amd64 are different than i386 resulting in
      different files on amd64 and i386.
      (LP #1934936)

 -- Steve Langasek <email address hidden> Tue, 24 Jan 2023 19:14:54 -0800

Due to this, syncpackage doesn't run.
The package can be sync'd next time it comes up.

[Steve Langasek (vorlon)]

My own testing prior to upload showed that there were differences in the dependencies, so I'm not sure why you got a different result.

[Michał Małoszewski (michal-maloszewski99)]

Checked the builds that seem to show they are good without delta vs the originally reported bug https://bugs.launchpad.net/ubuntu/+source/heimdal/+bug/1934936

Checking the non-lto PPA the two changelogs are now matching:

$ dpkg -x libwind0-heimdal_7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2_amd64.deb hd-amd64
$ dpkg -x libwind0-heimdal_7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2_i386.deb hd-i386
$ md5sum hd-amd64/usr/share/doc/libwind0-heimdal/changelog.Debian.gz hd-i386/usr/share/doc/libwind0-heimdal/changelog.Debian.gz
5dd57970afedcb4a75ece78e94109d17 hd-amd64/usr/share/doc/libwind0-heimdal/changelog.Debian.gz
5dd57970afedcb4a75ece78e94109d17 hd-i386/usr/share/doc/libwind0-heimdal/changelog.Debian.gz

Build logs:
https://launchpadlibrarian.net/646899573/buildlog_ubuntu-lunar-amd64.heimdal_7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2_BUILDING.txt.gz
https://launchpadlibrarian.net/646899827/buildlog_ubuntu-lunar-i386.heimdal_7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2_BUILDING.txt.gz

no more show "Searching for duplicated docs in dependency libroken18-heimdal... symlinking changelog.Debian.gz in" despite being built without LTO.

[Michał Małoszewski (michal-maloszewski99)]

Moreover here are my 2 PPAs that has been compared:

https://launchpad.net/~michal-maloszewski99/+archive/ubuntu/heimdal-test-ppa2-michal
https://launchpad.net/~michal-maloszewski99/+archive/ubuntu/test-heimdal-lto-michal-ppa1

[Christian Ehrhardt  (paelzer)]

Hmm - odd,
Michal has checked the files and the build logs as he outlined above.
And in addition to all those checks being done, if we just grep for all final dependencies and compare there is no difference (other than a glibc min version level).

I know names are not too helpful:
- ~ppa1 = LTO-off
- ~ppa2 = LTO-on

This compares amd64 vs i386 without LTO enabled

$ grep "Depends" buildlog_ubuntu-lunar-amd64.heimdal_7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2_BUILDING.txt > amd64.deps
$ grep "Depends" buildlog_ubuntu-lunar-i386.heimdal_7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2_BUILDING.txt > i386.deps
$ diff -Naur amd64.deps i386.deps
--- amd64.deps 2023-05-16 10:18:49.754334825 +0200
+++ i386.deps 2023-05-16 10:18:51.106344323 +0200
@@ -8,10 +8,10 @@
  Depends: debconf (>= 0.5.00) | debconf-2.0, heimdal-clients, krb5-config, lsb-base, openbsd-inetd | inet-superserver, libasn1-8-heimdal (>= 1.4.0+git20110226), libc6 (>= 2.34), libcap-ng0 (>= 0.7.9), libgssapi3-heimdal (>= 1.4.0+git20110226), libhcrypto5-heimdal (>= 1.4.0+git20110226), libhdb9-heimdal (>= 1.6~git20131117), libheimntlm0-heimdal (>= 1.4.0+git20110226), libkadm5srv8-heimdal (>= 7.8.git20221115.a6cf945+dfsg), libkdc2-heimdal (>= 1.4.0+git20110226), libkrb5-26-heimdal (>= 1.7~git20160418), libroken19-heimdal (>= 1.7~git20150920), libsl0-heimdal (>= 1.4.0+git20110226)
  Depends: comerr-dev, libasn1-8-heimdal (= 7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2), libgssapi3-heimdal (= 7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2), libhcrypto5-heimdal (= 7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2), libhdb9-heimdal (= 7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2), libheimbase1-heimdal (= 7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2), libhx509-5-heimdal (= 7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2), libkadm5clnt7-heimdal (= 7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2), libkadm5srv8-heimdal (= 7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2), libkafs0-heimdal (= 7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2), libkdc2-heimdal (= 7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2), libkrb5-26-heimdal (= 7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2), libwind0-heimdal (= 7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2), libotp0-heimdal (= 7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2), libsl0-heimdal (= 7.8.git20221117.28daf24+dfsg-1ubuntu1~ppa2), libc6 (>= 2.34), libcom-err2 (>= 1.43.9), libroken19-heimdal (>= 1.4.0+git20110226)
  Depends: krb5-config, openbsd-inetd | inet-superserver, libc6 (>= 2.34), libkrb5-26-heimdal (>= 1.4.0+git20110226), libroken19-heimdal (>= 1.4.0+git20110226)
- Depends: libc6 (>= 2.14), libcom-err2 (>= 1.43.9), libroken19-heimdal (>= 1.4.0+git20110226)
+ Depends: libc6 (>= 2.8), libcom-err2 (>= 1.43.9), libroken19-heimdal (>= 1.4.0+git20110226)
  Depends: libasn1-8-heimdal (>= 1.4.0+git20110226), libc6 (>= 2.34), libcom-err2 (>= 1.43.9), libhcrypto5-heimdal (>= 1.4.0+git20110226), libheimntlm0-heimdal (>= 1.4.0+git20110226), libkrb5-26-heimdal (>= 1.6~git20131117), libroken19-heimdal (>= 1.7~git20150920)
  Depends: libasn1-8-heimdal (>= 1.4.0+git20110226), libc6 (>= 2.36), libheimbase1-heimdal (>= 1.4.0+git20110226), libroken19-heimdal (>= 1.7~git20150920)
- Depends: libasn1-8-heimdal (>= 1.6~git201...

[Christian Ehrhardt  (paelzer)]

@Steve / @Vorlon

As outlined above we still can't see the diff in dependencies due to LTO.
But I'm sure you have seen it or you wouldn't have said so and we want to spot where in our work the mistake was ...
Therefore let me ask - was that a local build that you did or is that somewhere we could have a look at for comparison?

[Michał Małoszewski (michal-maloszewski99)]

Hi @Steve / @Vorlon

Could you help me identify where I might make a mistake in my investigation in Heimdal? I am writing that comment after a month again, because I would love to gain more knowledge and know where I was wrong.

Thank you in advance

[Michał Małoszewski (michal-maloszewski99)]

I wanted to make sure again if my reasoning was correct (I guess it was), so I did rebuild in 2 PPAs.
The ppc64el build failed, I did re-run the build and it failed again, so it is reproducible.

Due to FTBFS we need to keep disabling LTO.

New PPAs (named more explicitly, so now there is no problem distinguishing which is which):

https://launchpad.net/~michal-maloszewski99/+archive/ubuntu/ppa1-heimdal-test-lto-off
https://launchpad.net/~michal-maloszewski99/+archive/ubuntu/ppa2-heimdal-test-lto-on

To sum up, we had a reason to disable LTO; this reason is gone, so we wanted to make it a sync; now there is a new reason to keep disabling LTO.

I am going to set the status of the bug as a Won't Fix.

[Steve Langasek (vorlon)]

FTR, I did a local rebuild of heimdal 7.8.git20221117.28daf24+dfsg-3 on amd64 and see:

 Package: libwind0-heimdal
 Depends: libc6 (>= 2.14), libcom-err2 (>= 1.43.9)

# Older entries have been removed from this changelog.
# To read the complete changelog use `apt changelog libwind0-heimdal`.

And when disabling LTO:

 Depends: libc6 (>= 2.14), libcom-err2 (>= 1.43.9), libroken19-heimdal (>= 1.4.0+git20110226)

So I don't know why you're not seeing the same delta on amd64!