Re: Release major version number change causing upgrade problems

[Greg Troxel]

Carl Laufer  writes:

> If the upstream osmocom soversion is bumped, then all forks will eventually
> need to follow too. Because software packages will begin getting written
> for version 2, then those forks will need to update to be useful to those
> programs.

I think the real issue is what are forks doing and why.

There are basically two kinds:

  [things that aren't forks that github wrongly labels forks, which are
  just personal copies, either to have the bits, or as staging areas for
  proposing changes]

  a fork that intends to be compatible with a few features that are not
  yet in the main repo, or that have been rejected.  here, the release
  versions and soversion should track more or less exactly

  a fork that rejects the original repo, is renamed, and is its own
  thing

Then there are mirrors on different forges, often because some people
don't like other's forges (e.g., because github is proprietary, or
because !github is perceived as leaving out the used-to-monopoly crowd).

Often (not about this situation), I find that forks do not clearly state
that they are a fork, what the plan is and why, and what the status is.

I have found this community to be particularly hard to understand in
terms of fork status, relative to most others.  

Overall, a healthy ecosystem
So: for each fork, why does it exist, and is there a plan to fold it in?

Re: Release major version number change causing upgrade problems

[Андрей Родионов]

Hi guys,

FYI If the 2.x version of the package is installed, it will not be 
automatically replaced by version 0.x, even if 0.x is newer and 2.x is older. 
The APT package manager adheres to the semantic versioning scheme to identify 
the newer version.

Best Regards,
Andrey

> On 27 Mar 2024, at 00:11, Carl Laufer  wrote:
> 
> Hi Steve,
> 
> If the upstream osmocom soversion is bumped, then all forks will eventually 
> need to follow too. Because software packages will begin getting written for 
> version 2, then those forks will need to update to be useful to those 
> programs.
> 
> I believe the soversion should only be bumped when there have been breaking 
> non-backwards compatible API changes made. Nothing of that sort has happened 
> with the recent pushes.
> 
> So IMO I think reverting the soversion to 0 is the best solution, and that 
> should be pushed ASAP before the problem becomes more widespread. I just 
> talked to the dev of SDR++ and he agrees with this solution too. The distro's 
> that already picked it up probably should be notified that the soversion was 
> bumped by mistake and that it's been reverted.
> 
> Regards,
> Carl Laufer
> 
> 
> On Wed, Mar 27, 2024 at 9:15 AM Steve Markgraf  > wrote:
>> Hi,
>> 
>> the reason why the version was bumped is because there are several
>> forks of rtl-sdr that used version 0.8 and beyond, without changing the
>> library name. Many people requested the release of a new version, and to
>> avoid colissions of the version number with those forks, the major 
>> version was bumped to 2 as a 'leap forward'.
>> 
>> On 22.03.24 05:05, Carl Laufer wrote:
>> > To add to this here is a Twitter/X thread that the developer of SDR++ 
>> > has put out, regarding the issues he's seeing with this major version 
>> > number change.
>> > 
>> > https://twitter.com/ryzerth/status/1771016439681466697 
>> > 
>> 
>> 
>> Interestingly the last reply in that thread is from someone who
>> maintains one of those forks of rtl-sdr, and the first thing he did was 
>> change the version to 2.1 - so yeah, in the end it was useless.
>> 
>> We could probably revert the SOVERSION to 0, however, for those
>> distributions that already picked up the change this would be pretty
>> weird.
>> 
>> Regards,
>> Steve
>> 

Re: Release major version number change causing upgrade problems

[Carl Laufer]

Hi Steve,

If the upstream osmocom soversion is bumped, then all forks will eventually
need to follow too. Because software packages will begin getting written
for version 2, then those forks will need to update to be useful to those
programs.

I believe the soversion should only be bumped when there have been breaking
non-backwards compatible API changes made. Nothing of that sort has
happened with the recent pushes.

So IMO I think reverting the soversion to 0 is the best solution, and that
should be pushed ASAP before the problem becomes more widespread. I just
talked to the dev of SDR++ and he agrees with this solution too. The
distro's that already picked it up probably should be notified that the
soversion was bumped by mistake and that it's been reverted.

Regards,
Carl Laufer


On Wed, Mar 27, 2024 at 9:15 AM Steve Markgraf  wrote:

> Hi,
>
> the reason why the version was bumped is because there are several
> forks of rtl-sdr that used version 0.8 and beyond, without changing the
> library name. Many people requested the release of a new version, and to
> avoid colissions of the version number with those forks, the major
> version was bumped to 2 as a 'leap forward'.
>
> On 22.03.24 05:05, Carl Laufer wrote:
> > To add to this here is a Twitter/X thread that the developer of SDR++
> > has put out, regarding the issues he's seeing with this major version
> > number change.
> >
> > https://twitter.com/ryzerth/status/1771016439681466697
> > 
>
>
> Interestingly the last reply in that thread is from someone who
> maintains one of those forks of rtl-sdr, and the first thing he did was
> change the version to 2.1 - so yeah, in the end it was useless.
>
> We could probably revert the SOVERSION to 0, however, for those
> distributions that already picked up the change this would be pretty
> weird.
>
> Regards,
> Steve
>
>

Re: Release major version number change causing upgrade problems

[Steve Markgraf]

Hi,

the reason why the version was bumped is because there are several
forks of rtl-sdr that used version 0.8 and beyond, without changing the
library name. Many people requested the release of a new version, and to
avoid colissions of the version number with those forks, the major 
version was bumped to 2 as a 'leap forward'.


On 22.03.24 05:05, Carl Laufer wrote:
To add to this here is a Twitter/X thread that the developer of SDR++ 
has put out, regarding the issues he's seeing with this major version 
number change.


https://twitter.com/ryzerth/status/1771016439681466697 




Interestingly the last reply in that thread is from someone who
maintains one of those forks of rtl-sdr, and the first thing he did was 
change the version to 2.1 - so yeah, in the end it was useless.


We could probably revert the SOVERSION to 0, however, for those
distributions that already picked up the change this would be pretty
weird.

Regards,
Steve

Re: Release major version number change causing upgrade problems

[Carl Laufer]

To add to this here is a Twitter/X thread that the developer of SDR++ has
put out, regarding the issues he's seeing with this major version number
change.

https://twitter.com/ryzerth/status/1771016439681466697

Regards,
Carl


On Fri, Mar 22, 2024 at 3:19 PM Carl Laufer  wrote:

> Hi, I just saw that the major version number was recently bumped on the
> osmocom rtl-sdr.
>
> Unfortunately this seems to be causing upgrade problems because if a user
> updates the drivers to the new major version number (eg to get Blog V4
> support), the soname is now different, and this means that all previously
> installed software that uses librtlsdr will need to be recompiled from
> scratch, and software from the package manager will simply be broken.
>
> I'm not sure if the recent changes warrant a major version change and the
> problems that come with it?
>
> Regards,
> Carl
>