Libstellarsolver-dev dependency

[Eric Dejouhanet]

Hello,

So there were actually two issues, explained in issue 37.

The first issue was that the docker image used for building contained an
old release of libstellarsolver-dev, which pulled libstellarsolver during
its generation.
As a result, updating libstellarsolver-dev at the beginning of each build
never pulled libstellarsolver and its library, as the dependency was
already present.

This is now fixed in the CI recipe, but this situation reveals a second
issue that is maybe more serious.

Updating the libstellarsolver-dev package does not update the
libstellarsolver dependency, thus the end-user gets updated headers and old
libraries.
The minimum version of the libstellarsolver package should be mentioned in
the control script of the libstellarsolver-dev package, so that the latter
pulls the first on update.

Le sam. 7 nov. 2020 à 23:46, Robert Lancaster <rlancaste at gmail.com> a
écrit :

> Ok, so this is an issue with the linux distribution package and installing
> the libraries to the correct “libdir” on different linux systems. I
> definitely would not know how to fix that, but maybe Jasem does
>
>
> On Nov 7, 2020, at 4:56 PM, Eric Dejouhanet <eric.dejouhanet at gmail.com>
> wrote:
>
> That's it. If you really want to test you need to install the package. I
> can confirm installing from the build folder places files at the right
> location. Installing from the package does not.
>
> eric.dejouhanet at gmail.com - https://astronomy.dejouha.net
> *De:* rlancaste at gmail.com
> *Envoyé:* 7 novembre 2020 21:29
> *À:* mutlaqja at ikarustech.com
> *Cc:* eric.dejouhanet at gmail.com; kstars-devel at kde.org
> *Objet:* Re: Libstellarsolver-dev dependency
>
> I have myself seen some issues on some linux distributions with installing
> INDI drivers.  When I was making my AstroPi3 installation scripts, most of
> them worked fine, but there were a few that installed libraries to a lib
> directory where the INDI driver that depended on that library could not
> find it during building.  I eventually solved the issue by copying the
> libraries in the script to the other lib directory, but that is not the
> best solution.  The INDI drivers should be able to search the correct lib
> directory to find the libraries that were installed and the libraries
> should likewise install to the correct directory so that things that depend
> on them can find them.
>
> Is this a similar issue?
>
> On Nov 7, 2020, at 3:19 PM, Jasem Mutlaq <mutlaqja at ikarustech.com> wrote:
>
> Ubuntu here no issues. ALso on Windows/MacOS no issues.
>
> --
> Best Regards,
> Jasem Mutlaq
>
>
>
> On Sat, Nov 7, 2020 at 10:09 PM Eric Dejouhanet <eric.dejouhanet at gmail.com>
> wrote:
>
>> OK I will have a look at this then.
>> If it works on Windows, on Mac and on RPi then it is probably a Linux
>> distribution discrepancy. The CI uses an Ubuntu docker image. What are you
>> guys using?
>>
>> eric.dejouhanet at gmail.com - https://astronomy.dejouha.net
>>
>>
>>
>>           Message original
>>
>>
>> De: rlancaste at gmail.com
>> Envoyé: 7 novembre 2020 18:50
>> À: eric.dejouhanet at gmail.com
>> Cc: kstars-devel at kde.org
>> Objet: Re: Libstellarsolver-dev dependency
>>
>>
>> Hi Eric,  I am by no means an expert in this.  Originally I had the
>> library install to the machine's libdir and to /usr/lib because on some
>> systems it wasn’t getting found properly by other programs.  I was using the
>>  FindStellarSolver.cmake <http://findstellarsolver.cmake/> in kstars
>> which was based on the other ones that were already there.  It used
>> pkgconfig or an alternate method to find the library if that didn’t work.
>> I used the others as examples since I didn’t know the details on how to do
>> it.  This seemed to work pretty well.  Jasem said that we just needed to
>> install to the libdir and not /usr/lib, so we removed one of the installs.
>> He then said that there was some kind of craft issue with finding
>> StellarSolver on some windows machines, although I had tested it on Windows
>> and it seemed to work great.  He said that Hannah said that we should be
>> doing it a different “modern” way and not use the FindStellarSolver method
>> that we were using, so he set it up her way, both in KStars and
>> StellarSolver.  Possibly somewhere in those changes it stopped working for
>> some machines.  As far as Jasem and I can tell on our machines it works
>> fine.  If you know what the problem is and know how to fix it, please
>> enlighten us.
>>
>> Thanks,
>>
>> Rob
>>
>> > On Nov 7, 2020, at 12:40 PM, Eric Dejouhanet <eric.dejouhanet at gmail.com>
>> wrote:
>> >
>> >> You have the complete log? The latest libstellarsolver installs
>> libstellarsolver.so.1.5
>> >
>> > The log I pasted is the failure of the KStars cmake configuration. It
>> > is still current as of today.
>> > The latest libstellarsolver may install that library, but it is indeed
>> > not put at the right location, and cannot be found.
>> > This is for arch amd64.
>> >
>> > --
>> > -- eric.dejouhanet at gmail.com - https://astronomy.dejouha.net
>>
>>
>
>
>

-- 
-- eric.dejouhanet at gmail.com - https://astronomy.dejouha.net
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kstars-devel/attachments/20201114/09b6916d/attachment.htm>