Re: Request for Rawtran packaging review
Hi Ole,
On Tue, 2016-04-05 at 08:53 +0200, Ole Streicher wrote:
> Filip Hroch <hroch@physics.muni.cz> writes:
> > I added both kinds (build and CI) of the tests. Now, I see that the
> > idea wasn't so good because there are no small CR2 files due to the
> > race of manufacturers about mega-pixels.
>
> Probably setting the resolution in the camera settings to some very low
> number doesn't help since the raw file is saved before any resolution
> change is done, right?
>
> You could (if you know the format well enough) try to create a synthetic
> raw file, or you could set most of the pixels to zero (that could make
> it better compressible). Or just live with the 8 MB tar file, which is
> luckily not a problem anymore today.
>
There are no a standard format of raw photos. Every manufacturer has
its own (frequently, a few) standards developed. This is the cruital
problem with raw photos. Therefore only one correct format is the
one created by a camera. That's dcraw is so great piece of sofware,
Mr. Coffin knows (perhaps due to a background informations) detailed
structure of all the implemented formats.
> One point here is however that the CI test as you specified it does not
> work: the source directory is readonly during the tests. And the build
> environment (namely gcc, make and such) is not installed by default
> there, it should be specified in the Depends fields of
> debian/tests/control.
>
It is possible that the source directory is read only. But the
test directory must be read write to be able create test files
(.fits in this case) so I understand that all contents of the
package is copied to the rw directory and tests are run within it.
I test the feature with reccomended command:
adt-run --output-dir /tmp/output-dir ./ --- null
passing -s (--shell) option which provides the development
environment. It shows that the build tests works clearly.
I see only one problem: I omit of Build-depends
because all the tools (gcc, make) are installed on
my machine.
Or Am I wrong?
> Since your package already contains an executable: wouldn't it be better
> in the CI test just to create the fits file with rawtran, and then check
> some headers (and, if you know it, the content)? You could use f.e. the
> "fitshdr" command from wcslib-tools for that. Or astropy.
>
Yes. Astropy is the solution in case of read only tests. But I prefers
the more simple and clear way via C. The upstream needs only plain C
compiler and any other dependence can importantly complicate life of
potential users. On the other side, it is not a good idea to maintain
two identical test suites in C and Python.
FH
--
F. Hroch e-mail, jabber: hroch@physics.muni.cz, tel.: +420549494470
Dept. of theor. physics and astrophysics, MU Brno, Kotlarska 2,CZ-611 37
Reply to: