Re: [Outreachy] You can use "Team upload" (Was: r22074 - in trunk/packages/conservation-code/trunk/debian: . tests)
[CCing the list for more input - I'll be not reliably available online today
and tomorrow]
Good morning Tanya,
On Thu, Jun 02, 2016 at 08:37:36AM +0300, merlettaia wrote:
> The test I wrote is unfinished - I had an error message when I tried to run
> this simple test locally, now I'm stuck, and can't decide how to fix.
> I think I should provide some details - maybe this error unreproducible,
> and appears only on my VM, I don't know. And some questions - maybe you
> could help me find the best way to fix them. There are several related
> problems with this package:
>
> Problem 1. When I run adt-run, this error appears:
> "...
> s_matrix = read_scoring_matrix(s_matrix_file)
> File "/tmp/autopkgtest.ASxyOi/deps/usr/bin/score_conservation", line 547,
> in read_scoring_matrix
> return identity(20)
> NameError: global name 'identity' is not defined
> "
> - it appears because package numpy, where identity method is located, is
> imported in conditional statement, and is not imported globally in module.
Hmmm, I admit I do not understand this since I called the script from my
local machine (which would be nice to work as well in any case).
> When alignment matrix, which by default is searched at matrix/blosum62.bla
> (directory is relative to script),
This is true in the source directory layout. However, I think we should
make this working even outside this directory layout to serve as a
sensible example for users. While adt-run is running inside the
unpackaged source and we can assume what you said I usually install the
test script as example for users in /usr/share/doc/<packagename> to
learn how to use the program.
Since the matrix files are installed to
/usr/share/conservation-code/matrix
conservation-code should find these files there.
> can't be found at that location, this
> identity method is used to get some 'default' alignment matrix.
> The fix for this error must be simple - to add import statement before
> calling identity method (fix upstream code).
Sounds sensible. You might like to create a quilt patch (if specifying
/usr/share/conservation-code/matrix does not help).
> Now this error shows that a package is installed incorrectly, and some
> related files can't be found.
>
> I think I have to fix upstream and to add separate test to ensure all goes
> fine with default parameters. Do you agree, or is it redundant?
Agreed.
> Problem 2. When I run svn-buildpackage with current debian/install file,
> these messages appear:
> "...
> /data/src/build-area/conservation-code-20110309.0/debian/install: 1:
> /data/src/build-area/conservation-code-20110309.0/debian/install:
> distributions: not found
> /data/src/build-area/conservation-code-20110309.0/debian/install: 2:
> /data/src/build-area/conservation-code-20110309.0/debian/install: matrix:
> not found
> /data/src/build-area/conservation-code-20110309.0/debian/install: 3:
> /data/src/build-area/conservation-code-20110309.0/debian/install:
> score_conservation: not found
> ...
> "
Are you running it from the conservation_code directory? I personally
do not use svn-buildpackage but rather pdebuild (to build in pbuilder or
simply debuild to run on the local machine. So the proper sequence to
build is
apt-get source conservation-code
cd conservation-code-20110309.0
# copy your changes from svn to debian/ dir but
# cd back to this dir
pdebuild
# or debuild
What you wrote sounds like starting in wrong dir.
> For the reasons unknown, data from matrix and distribution folders and
> score_conservation file can't be found.
I can not reproduce this.
> For now I think it's some strange
> problem with VM. Do this messages appear when you run svn-buildpackage? Or
> am I doing something wrong?
I do not think that the VM is to be blamed for this kind of error.
> Problem 3 - Combines problems 1 and 2.
> While installing, contents of the score_conservation.py (which by default
> uses matrix/blosum62.bla) is copied to score_conservation (which by default
> uses matrix/blosum62.bla), and this files goes to usr/bin/, when default
> alignment matrix file goes in package folder. Thus, when we run
> score_conservation with default alignment matrix, error message described
> as Problem 1 appears.
That's why I said we probably need to adapt the path to matrix dir inside
the code.
While beeing at it: The current packaging method is wrong since Python
code should go to /usr/share/<packagename> to enable pybuild create .pyc
files. It might be (or not) that in this case the matrix files will be
found. Please wait until you see my commit for this.
> For me it seems to be the best solution to use a symlink named
> score_conservation to score_conservation.py instead of copying that file -
> to keep python script in package folder to avoid problems with default
> parameters (it can also require small fix in upstream). Do you think it is
> ok or there can be better solution?
Yes, it is - also for the reasons I mentioned above.
> Sorry if I'm unclear or too verbose :(
You are pretty clear. ;-)
> And thank you for remarks!
You are welcome
Andreas.
--
http://fam-tille.de
Reply to: