Hi Pranav,
On Sat, May 09, 2020 at 09:54:56PM +0530, Pranav Ballaney wrote:
> I've added autopkgtests to cgview. Since the output is an image, I haven't
> compared it to a reference - is that alright?
Its definitely better than nothing.
> If not, how do we usually test packages like this? I tried running a diff
> but that reported the files to be different, even though they were visually
> the same.
I somehow thought there are some image comparison packages inside Debian.
A short search has lead me to
https://softwarerecs.stackexchange.com/questions/9774/command-line-tool-to-check-whether-two-images-are-exactly-the-same-graphically
which recommends
identify -quiet -format "%#" images...
> Please take a look and let me know if any additional steps are required.
> https://salsa.debian.org/med-team/cgview
Please note that I used xz compression on the xml data to save some
space also on developers machines. I've also called routine-update -f
to update the packaging. I do not think that the image comparison is
mandatory here - its way better to test at all than to make it perfect.
But if you want to give identify a try that would be fine.
Okay then let's leave the test here. Thanks for the compression, I'll keep that in mind for future tests.
However, we have a remaining problem in my pbuilder chroot:
autopkgtest [19:32:42]: test run-unit-test: [-----------------------
run-detectors: unable to find an interpreter for /usr/bin/cgview
autopkgtest [19:32:43]: test run-unit-test: -----------------------]
autopkgtest [19:32:43]: test run-unit-test: - - - - - - - - - - results - - - - - - - - - -
run-unit-test FAIL non-zero exit status 2
This is weird - all tests pass in my chroot. Could it be like the issue we had with loki a few months ago?
It worked perfectly on Liubov's machine too.
@Nilesh, could you please take a look here?
It somehow sounds similar to this thread:
https://lists.debian.org/debian-java/2016/04/msg00098.html
I'd recommend to check with debian-java@lists.debian.org . In
the case of artfastqgenerator I've circumvented the problem
a wrapper script.
Let's let Nilesh try it once, and if the problem persists, we'll take it up with the Java team.
Regards,
Pranav