[Ideas] Improve Version Information for Source Builds

[tuhaihe]

Description

Currently, when Cloudberry is built from a source zip file (not cloned via git), the version information displayed is not matched the right info, always like this:

PostgreSQL 14.4 (Cloudberry Database 1.0.0 build dev) on x86_64-pc-linux-gnu...

This is not user-friendly for those compiling and installing Cloudberry from a downloaded source zip files. It makes it difficult for users to report issues, perform version upgrades, or confirm the exact build version, a concern raised multiple times across community channels in the past months.

I noticed that PostgreSQL addresses this by defining the version information explicitly in cases where git metadata is unavailable, as seen in this commit.

Would it be possible to adopt a similar approach in Cloudberry for non-git builds? I believe this could greatly improve version traceability. Would like to have your ideas.

Use case/motivation

No response

Related issues

No response

Are you willing to submit a PR?

• Yes I am willing to submit a PR!

[my-ship-it]

Is there a bug for versioning scripts? Why it is always 1.0.0 for Cloudberry Database.
How about follow format below for debug/release?

PostgreSQL 14.4 (Cloudberry Database 1.5.2+4859142 build dev) on x86_64-pc-linux-gnu, compiled by gcc (GCC) 9.5.0, 64-bit compiled on Dec 12 2024 22:18:23
or
PostgreSQL 14.4 (Cloudberry Database 1.5.2+4859142 build release) on x86_64-pc-linux-gnu, compiled by gcc (GCC) 9.5.0, 64-bit compiled on Dec 12 2024 22:18:23

Where 4859142 is the commit sha1 so that we could track down to commit point.

[tuhaihe]

Is there a bug for versioning scripts? Why it is always 1.0.0 for Cloudberry Database. How about follow format below for debug/release?

I guess there is no mechanism to detect the version information for the downloaded source tar files. It would be better if an engineer could help debug it.

PostgreSQL 14.4 (Cloudberry Database 1.5.2+4859142 build dev) on x86_64-pc-linux-gnu, compiled by gcc (GCC) 9.5.0, 64-bit compiled on Dec 12 2024 22:18:23
or
PostgreSQL 14.4 (Cloudberry Database 1.5.2+4859142 build release) on x86_64-pc-linux-gnu, compiled by gcc (GCC) 9.5.0, 64-bit compiled on Dec 12 2024 22:18:23

Where 4859142 is the commit sha1 so that we could track down to commit point.

Agree. If we build the Cloudberry by git clone the source, will return like this.

[my-ship-it]

There is a script called https://github.com/apache/cloudberry/blob/main/getversion to do this. @edespino Could you please double check with it, thanks.

[tuhaihe]

Hey, I found we need to change the string [1.0.0] in this file for each release, like [2.0.0-Aplha], [2.0.0-Beta]:

cloudberry/configure.ac

Line 23 in dd5aa20

AC_INIT([Cloudberry Database], [1.0.0], [info@cloudberrydb.org], [], [https://www.cloudberrydb.org/])

Also need to generate a new configure file, and then return the right release number when building from the source tar files.

Also, we can create one file VERSION, and input the version number, then it also can work.

In summary, for a lower workforce, we can set [1.0.0] to [main] in the configure.ac file—keep it unchanged in the future. We can also create a new text file, VERSION, and update it when a new version is released. It will be easier to change one text file than to regenerate the configure file by the autoconf command.

[tuhaihe]

I can help create one PR to fix this.

[tuhaihe]

One PR has been created for this discussion: #929. Welcome to have your comments.