[arch-general] changelogs (was Re: postgresql 9.3 -> 9.4)

Georg Altmann george at george-net.de
Thu Jan 29 16:06:43 UTC 2015

Hash: SHA1

On 29.01.2015 16:33, Lukas Fleischer wrote:
> On Thu, 29 Jan 2015 at 16:08:02, Carl Schaefer wrote:
>> The thread about the postgresql update reminded me of one of the 
>> few things about Ubuntu that I miss: package updates usually 
>> included a useful changelog entry describing what was fixed 
>> and/or new.  Perhaps I assume too much, but I imagine Arch 
>> package maintainers would generally be aware of the changes in
>> an update they're making, and so it wouldn't be much additional
>> work to include that information in a changelog entry.
> It actually *is* much additional work. Bumping the package version
>  usually takes ~5 seconds. That does not include the time to build 
> and test the package but we don't need to watch the build process 
> and packages are often tested by using them in production. So, 
> compared to those five seconds, looking up every single change and 
> coming up with a change log is a lot of time -- and it usually 
> doesn't pay off, apart from corner cases. Please examine the 
> mailing lists, this story has been discussed several times
> already. Keeping bureaucracy to a minimum is one of the reasons we
> can provide updates much faster than other distributions.

I understand the effort the packaging requires. Writing changelogs may
be to much work and this is not what I propose. On the other hand,
upgrades that break stuff are seldom. As is the case for postgresql.
This happens every few month or so.

On 29.01.2015 14:22, Bardur Arantsson wrote:
> If the problem here is that it would be a chore to do this for 
> maintainers for every X.Y -> X.(Y+1) upgrade, then maybe Arch 
> package descriptions could grow a field or flag to handle such 
> things semi-automatically? Maybe something as simple as "if the 
> version number is about to change in *this way*, then warn loudly 
> using *this message*".

Wouldn't that be a sensible way? The increased overhead for the
maintainer would be to tick a flag in addition to the version bump.
In the case of postgresql this would be a as simple as

if (oldMajor < newMajor ||
  ((oldMajor == newMajor) && (oldMinor < newMinor)) {

Of course the condition would have to be serialized in the package
meta-data some way. I have only very limited knowledge on the pacman
internals. Maybe someone can come up with an estimate how big the
effort would be to implement this.

> While I agree that warnings and front page news should be given 
> where appropriate, I cannot comment on PostgreSQL, which I don't 
> use. As it seems to be a similar process on every major update and 
> there even seems to be a script to warn you, I don't see any need 
> for another notice, though. As a database administrator, you
> should be aware of what happens when you update the DBMS. Maybe
> some post-upgrade message would be helpful...

Agreed -- it's just that I am not a DBMS admin in this case. This was
on my personal computer where I can hardly spent the time to look up
every package version change. In times where remotely exploitable
security flaws turn up almost daily, this is just not acceptable.
So, I have to trust the packaging system to some degree. Of course one
should check Arch announcements on the website and maybe follow the
mailing lists. There wasn't any notice for postgresql in this case.

Version: GnuPG v1


More information about the arch-general mailing list