A few questions about rpmdev-bumpspec tool

[Qiyu Yan]

Hello all,

I have some problem with rpmdev-bumpspec recently.

In the latest version of rpmdevtools, rpmdev-bumpspec has changed to
use time+date in the changelog it generates[1], while the packaging
guidelines have not been updated accordingly[2], should the guideline
be updated to the rpmdev-bumpspec change?

I am packaging fcitx5 using forge macros, and upstream have never
tagged a version, in this case, I am packaging like this [3] (The
snapshot dates and git short commit hashesin changelog is manually
added). With this spec file, I noticed that when I try to use
rpmdev-bumpspec to generate a changelog, it will give things like this
[4].

You can see that, in case of using forge, rpmdev-bumpspec can't
include either snapshot dates nor git short commit hashes, will this
be fine (and we can ignore the warning from rpmlint when ran on the
built packages, and start the review process) or I should always
manually include snapshot dates and git commit hashes in the
changelog. Or I should wait for this change [5] to be done and ignore
all changelog things? (and submit for review then?)

Thanks.

[1]: https://pagure.io/rpmdevtools/c/d205ad9cfc4b7123acd573e028f8c4521ec79300
[2]: https://docs.fedoraproject.org/en-US/packaging-guidelines/#changelogs
[3]: https://github.com/karuboniru/fcitx5-fedora/blob/master/fcitx5/fcitx5.spec
[4]:
https://github.com/karuboniru/fcitx5-fedora/blob/972fd2e2e84e6ca136a9c5f4...
[5]:
https://fedoraproject.org/wiki/Changes/Patches_in_Forge_macros_-_Auto_mac...
------
And the snapshot dates generated on my machine and copr can be
different, I think this is related to time zone (I am in UTC+8), I
don't think it is a bug, but I hope this will be improved.
-- 
Best regards,
Qiyu Yan

[Michael Schwendt]

On Thu, 13 Aug 2020 18:41:23 +0800, Qiyu Yan wrote:

>  In the latest version of rpmdevtools, rpmdev-bumpspec has changed to
>  use time+date in the changelog it generates[1], while the packaging
>  guidelines have not been updated accordingly[2], should the guideline
>  be updated to the rpmdev-bumpspec change? 
The git commit message for that change is vague, since it only refers to
an RPM version but doesn't sum up what the goal of this change is.

>  I am packaging fcitx5 using forge macros, and upstream have never
>  tagged a version, in this case, I am packaging like this [3] (The
>  snapshot dates and git short commit hashesin changelog is manually
>  added). With this spec file, I noticed that when I try to use
>  rpmdev-bumpspec to generate a changelog, it will give things like this
>  [4]. 
The forge macros you use mess with %dist, which is highly questionable.

The rpmdev-bumpspec script itself doesn't evaluate any RPM macros. It only
recognizes a variety of version/release schemes used by Fedora.

For the %changelog comment, it relies on an "rpm" command-line call in
order to determine the full E:VR for the %changelog entry. Since %dist is
not to be included in %changelog comments, %dist gets undefined, but then
the %forge macro stuff is lost.

As a side-note: The E:V-R details right of the email address in changelog
comments are not everyone's cup of tea. They are not mission critical but
optional. If truncated, they don't break the rpmbuild. In your case, V-R
is complete and accurate. The left most-significant part of %release is
included in the V-R and is sufficient, and while the git commit hash is
missing, it is just noise in changelog comments.

[Fabio Valentini]

On Thu, Aug 13, 2020 at 12:42 PM Qiyu Yan <yanqiyu(a)fedoraproject.org&gt; wrote:
>  Hello all,
>
>  I have some problem with rpmdev-bumpspec recently.
>
>  In the latest version of rpmdevtools, rpmdev-bumpspec has changed to
>  use time+date in the changelog it generates[1], while the packaging
>  guidelines have not been updated accordingly[2], should the guideline
>  be updated to the rpmdev-bumpspec change?
>
>  I am packaging fcitx5 using forge macros, and upstream have never
>  tagged a version, in this case, I am packaging like this [3] (The
>  snapshot dates and git short commit hashesin changelog is manually
>  added). With this spec file, I noticed that when I try to use
>  rpmdev-bumpspec to generate a changelog, it will give things like this
>  [4].
>
>  You can see that, in case of using forge, rpmdev-bumpspec can't
>  include either snapshot dates nor git short commit hashes, will this
>  be fine (and we can ignore the warning from rpmlint when ran on the
>  built packages, and start the review process) or I should always
>  manually include snapshot dates and git commit hashes in the
>  changelog. Or I should wait for this change [5] to be done and ignore
>  all changelog things? (and submit for review then?)
>
>  Thanks. 
You're not wrong, rpmdev-bumpspec is not completely compatible with
the forge macros, especially in the snapshot packaging case (because
the forge macros mess with the value of %dist, as mentioned in the
other response).

While rpmdev-bump will produce changelog entries with "wrong" (or
incomplete) EVR information, SRPM and RPM builds will have the correct
Release set, so it's not a "big deal". For my packages that suffer
from this (e.g. some Go packages that internally use the forge
macros), I tend to manually "fix" the version-release in the changelog
entries after running rpmdev-bumpspec.

Fabio

[Vít Ondruch]

Dne 13. 08. 20 v 12:41 Qiyu Yan napsal(a):
>  Hello all,
>
>  I have some problem with rpmdev-bumpspec recently.
>
>  In the latest version of rpmdevtools, rpmdev-bumpspec has changed to
>  use time+date in the changelog it generates[1], while the packaging
>  guidelines have not been updated accordingly[2], should the guideline
>  be updated to the rpmdev-bumpspec change? 

I don't like this change and asked revert:


https://pagure.io/rpmdevtools/issue/63


Vít


>  I am packaging fcitx5 using forge macros, and upstream have never
>  tagged a version, in this case, I am packaging like this [3] (The
>  snapshot dates and git short commit hashesin changelog is manually
>  added). With this spec file, I noticed that when I try to use
>  rpmdev-bumpspec to generate a changelog, it will give things like this
>  [4].
>
>  You can see that, in case of using forge, rpmdev-bumpspec can't
>  include either snapshot dates nor git short commit hashes, will this
>  be fine (and we can ignore the warning from rpmlint when ran on the
>  built packages, and start the review process) or I should always
>  manually include snapshot dates and git commit hashes in the
>  changelog. Or I should wait for this change [5] to be done and ignore
>  all changelog things? (and submit for review then?)
>
>  Thanks.
>
>  [1]: https://pagure.io/rpmdevtools/c/d205ad9cfc4b7123acd573e028f8c4521ec79300
>  [2]: https://docs.fedoraproject.org/en-US/packaging-guidelines/#changelogs
>  [3]: https://github.com/karuboniru/fcitx5-fedora/blob/master/fcitx5/fcitx5.spec
>  [4]:
> https://github.com/karuboniru/fcitx5-fedora/blob/972fd2e2e84e6ca136a9c5f4...
>  [5]:
> https://fedoraproject.org/wiki/Changes/Patches_in_Forge_macros_-_Auto_mac...
>  ------
>  And the snapshot dates generated on my machine and copr can be
>  different, I think this is related to time zone (I am in UTC+8), I
>  don't think it is a bug, but I hope this will be improved.