fun with lensfun

[Carl von Einem]

Hi,

just by chance I noticed that the Lensfun project doesn't mention Hugin
in their list of applications using their library:
http://lensfun.sourceforge.net/usage/

So I tried to make them aware of what I thought was just a little flaw
but I learned that the lensfun developers _insist_ that Hugin doesn't
use Lensfun which I think is funny. See
<http://sourceforge.net/p/lensfun/mailman/message/33389420/>.

So... is there some sort of hugin roadmap that either drops or includes
lensfun? They just don't believe me.

Cheers,
Carl

[T. Modes]

Hi Carl,

Am Donnerstag, 12. Februar 2015 16:07:50 UTC+1 schrieb zarl:

So... is there some sort of hugin roadmap that either drops or includes
lensfun? They just don't believe me.

In the default branch lensfun support was dropped and replaced with an own implementation.
The cause for this change is the "instability" of the lensfun library. The API is in continuous flow, features are added, renamed and maybe deleted.
And also the behaviour of some functions changes between different releases, and this without hint in the release notes or documentation. (For a developer this is more difficult to debug. You have to check the return values of each lensfun function, and this for each release.)
For a library which is mainly used by other programs (and rarely stand alone) I find is very strange, but this is only my opinion.
I don't have the time to change Hugins code for each lensfun release, otherwise the lensfun library does not work correctly. So I decided to remove lensfun from Hugin.

Thomas

[Carl von Einem]

Hi Thomas!

T. Modes wrote on 12.02.15 17:55:

Thanks for the clarification, that decision sounds very reasonable to
me. No wonder the lensfun developers only respond with what I recognise
as fuzzy phrases and avoid mentioning an application that will stop
using their library in the next release... ;-)

Cheers,
Carl

[Torsten Bronger]

Hallöchen!

Meanwhile, this thread was cited elsewhere, which thwarts my
original intention of ignoring it.

T. Modes writes:

> [...]

>
> The cause for this change is the "instability" of the lensfun
> library. The API is in continuous flow, features are added,
> renamed and maybe deleted. And also the behaviour of some
> functions changes between different releases, and this without
> hint in the release notes or documentation. (For a developer this
> is more difficult to debug. You have to check the return values of
> each lensfun function, and this for each release.) For a library
> which is mainly used by other programs (and rarely stand alone) I
> find is very strange, but this is only my opinion. I don't have
> the time to change Hugins code for each lensfun release, otherwise
> the lensfun library does not work correctly.

The lensfun library has a stable and well-documented API. If we
make changes, we follow the established procedures of obsolescence.
And even that has not been necessary so far. Moreover, we address
bugs very quickly.

Other programs like Darktable, Digikam, UFRaw, and rawstudio use
lensfun successfully, and never have complained about lensfun being
a moving target. So, for the sake of fairness to a fellow open
source project, please do not post such discrediting and vague
statements without evidence.

Tschö,
Torsten.

--
Torsten Bronger Jabber ID: torsten...@jabber.rwth-aachen.de
or http://bronger-jmp.appspot.com

[T. Modes]

Am Donnerstag, 19. Februar 2015 21:18:10 UTC+1 schrieb Torsten Bronger:

The lensfun library has a stable and well-documented API.  If we
make changes, we follow the established procedures of obsolescence.
And even that has not been necessary so far. 

LOL, from the changelog

Changelog for release 0.3.0

<snip>

• Replaced field-of-view data with real-focal-length data; marked the FOV API deprecated

This function was introduced in 0.2.6 and is now already deprecated.
No further comment.

Just one issue, the other one can everybody see in repository.

[Torsten Bronger]

Hallöchen!

T. Modes writes:

> Am Donnerstag, 19. Februar 2015 21:18:10 UTC+1 schrieb Torsten Bronger:
>
>> The lensfun library has a stable and well-documented API. If we
>> make changes, we follow the established procedures of
>> obsolescence. And even that has not been necessary so far.
>
> LOL, from the changelog
> Changelog for release 0.3.0
>
> <snip>
>

> - Replaced field-of-view data with real-focal-length data;

> marked the FOV API deprecated

Good example. This is exactly what I mean with "established
procedures": The exposed behaviour has not changed. The old
functions are there and fully working. We will not remove them
before the last significant vendor stops shipping programs using the
functions. Until then, there are directives in the code telling the
compiler to print warnings.

Thus, this is in accordance with best practices of software
engineering.

> This function was introduced in 0.2.6 and is now already
> deprecated. No further comment.

Well, you introduced it yourself for the Hugin integration of
lensfun, and Hugin has been its only user. And then Hugin departed.

> Just one issue,

... which was initiated after your removal of Lensfun from Hugin.
So, the only "issue" you present is one you was not even affected
by.

> the other one can everybody see in repository.

*You* made the allegation, so *you* must give evidence if you want
it to be valid. (You may consider lensfun's public mailing list for
that. In this case, link from here to there.)

[T. Modes]

Am Freitag, 20. Februar 2015 18:03:37 UTC+1 schrieb Torsten Bronger:

... which was initiated after your removal of Lensfun from Hugin.
So, the only "issue" you present is one you was not even affected
by.

*You* marked the fov api as deprecated on 13.04.2014 (http://sourceforge.net/p/lensfun/code/ci/d7464a97150b9267927ac32ea5fd0dce4eba98c9/)
When it is deprecated it can be removed in each following version without further notice.
So after *this* I removed the lensfun code from Hugin ( changeset 02283d5b9820bd73362e53c8ad9b6de59cb14465, 30.04.2014).

*You* mix up the order. Don't blame other for your actions.

[Torsten Bronger]

Hallöchen!

T. Modes writes:

> Am Freitag, 20. Februar 2015 18:03:37 UTC+1 schrieb Torsten Bronger:
>
>> ... which was initiated after your removal of Lensfun from Hugin.
>> So, the only "issue" you present is one you was not even affected
>> by.
>
> *You* marked the fov api as deprecated on 13.04.2014
> (http://sourceforge.net/p/lensfun/code/ci/d7464a97150b9267927ac32ea5fd0dce4eba98c9/)

I did so in my own branch "focal" (which still is on sourceforge),
*not* in "master".

On Apr 10th, I wrote in an email to the maintainer of lensfun (which
I am not) about "my plan" to mark the FOV API deprecated, "while
keeping it". He did not answer. I cannot decide such things, so, I
hacked away for another two weeks or so and then stopped working on
"focal".

On Aug 27th, I wrote an email to the maintainer that Hugin did not
use Lensfun anymore.

On Aug 30th, he merged the branch "focal" into "master". Thus, not
before this date the lensfun FOV API is on the deprecation path, in
the sense I explained in my previous posting.

> When it is deprecated it can be removed in each following version
> without further notice. So after *this* I removed the lensfun
> code from Hugin (changeset
> 02283d5b9820bd73362e53c8ad9b6de59cb14465, 30.04.2014).
>
> *You* mix up the order. Don't blame other for your actions.

You mixed up the branches.