On Thu, Jun 27, 2019 at 11:35:06AM +0100, Saira Hussain wrote:
after a lot of valuable help from Andreas, fixing syntax errors and looking
many hours through databases and gathering the important files (as well as
converting them) the ncbi-epcr is passing now the tests.
I can confirm that the tests are working now and I'll upload the
package. Congratulation to your first package in this Outreachy project
(well, it is really the "first" if we consider glam2 which you did in the
application phase as package number zero ;-) ).
What would be the next steps before I pick up the next bug? I have already
commited to master my latest changes and updated dch (using the gbp-dch, any
other recommendations?)
That's fine so far. I removed the "set -x" line which was helpful for
debugging but creates only noise in production. Further hint: For
debugging purposes it helps to comment out the line
trap "rm -rf ${AUTOPKGTEST_TMP}" 0 INT QUIT ABRT PIPE TERM
in run-unit-test. Thus you can see what files are available in
/tmp/${pkg}-test.XXXXXX which is sometimes helpful.
Regarding the next package: You can pick from the open bugs freely.
You can also do what I'm doing to know what packages would profit
most from a test. I've written an SQL query[1] for UDD (you need a
local psql client to run it) and I'm using it like this:
$ 0-missing-autopkgtest.sh d-m ebian-med bio bio-dev | grep -v -e autopkgtest -e "interface::x11" -e "uitoolkit::" debian-med-bio-tests.oout | head -n30
source | vote | testsuite | tags
------------------------------------------+------+-------------------------------------+------------------------------------------------------------------------
jmol | 86 | |
htsjdk | 30 | |
mcl | 27 | |
ngs-sdk | 14 | |
hyphy | 13 | |
biojava-live | 12 | |
jebl2 | 12 | |
librg-utils-perl | 10 | |
tree-puzzle | 10 | |
adun.app | 9 | |
mgltools-networkeditor | 9 | |
ncbi-seg | 9 | |
aevol | 8 | |
cdbfasta | 8 | |
diamond-aligner | 8 | |
epcr | 8 | |
norsnet | 8 | |
norsp | 8 | |
sibsim4 | 8 | |
staden | 8 | |
fastlink | 7 | |
freecontact | 7 | |
king-probe | 7 | |
loki | 7 | |
mgltools-pybabel | 7 | |
ncbi-entrez-direct | 7 | |
neobio | 7 | |
ngs-sdk | 7 | |
If you redirect
0-missing-autopkgtest.sh d-m ebian-med bio bio-dev
to some file you see a list of all Debian Med packages sorted by user numbers.
The grep filter after this removes all lines containing autopkgtest (these
packages are done as well as those that have some graphical interface (and
its usually hard to write a test for these (but may be some of these can be
called also with command line options? I don't know ...) So you end up with
a sorted list of 30 (just take the 30 important ones to pick one for you)
packages.
Looking at the list above here is a clear warning for
1. jmol
That's a huge one, outdated version and *really* hard to update
due to new Java dependencies. I never recommened it to newcomers
so leave it as is for the moment
2. htsjdk (#920588)
Might be an interesting target for someone who is *really* fluent
with Java and has *experience* with this toolkit. If you do not
subscribe to both features, just leave it for (way) later.
I'd strongly recommend to pick one from that list that rings a bell in
you what you might have used or intend to use or so. Also tell me
please in advance which one you think you have found a valuable target.
I checked the open bugs (clonalframe #891730, mipe #890788, tree-puzzle
#890790, hyphy #909707, loki #909708, altree #909710, fastlink #909713,
idba #909716) and it turns out that part of these have now lower number
of users or other packages just got more users. The bugs have no higher
importance than the table above - these just reflect the usage status
at the time of filing those bugs.
In short: pick a random one and tell me your decision before you invest
to much time into it. I might have additional hints.