[GSoC] Help needed for verifying cd-hit test (Was: [GSoC] ncbi-blast+ and htslib uploaded)
Hi Canberk,
just a general remark: It would help if you change the subject of your
mails if it does not fit the content any more.
On Thu, Aug 11, 2016 at 08:10:30AM +0300, Canberk Koç wrote:
> I commit Cd-Hit according to it's manual but these example codes did not
> work also there is no command like cd-hit when i install the program.
Well, if you do
dpkg -L cd-hit | grep bin
you can see what binaries are installed by the package. I admit I have
no idea at all why the links in /usr/bin are droping the '-' so cd-hit
becomes cdhit in some cases and in others it don't. Since no other
maintainer contributed to this package I need to blame myself for this
decision - but it seems somehow suspicious. Is there anybody who would
expect to find a /usr/bin/cdhit executable in any scripts or should we
rather rename to cd-hit to fit the upstream binary.
> So as
> you say i commit it with errors in test so we can figure it our together .
While I confirm that this is a good idea my expectation would be that
you exactly tell the problem here and not let me find out what might be
the problem and post it on behalf of yours here. I admit I do not
expect that here on the list are so many cd-hit users. Perhaps it might
be necessary to contact the upstream author about this. From my point
of view this kind of research belongs to your tasks in the GSoC project.
Is any user of this list able to make some sense out of this:
.../cd-hit/debian/tests.cd-hit$ cd-hit-para -i example.fa -o db90 -c 0.9 -n 5 -M 16000 –d 0 -T 0
no host at /usr/bin/cd-hit-para line 97.
(The exmaple.fa is in the Git repository but to reproduce the error any
*.fa might do.)
The issue might result from the attempt to query other computers of a
cluster but there is no such cluster configured. I also noticed that
the doc says:
[CD-HIT-PARA is no longer supported, since the multi-threaded cd-hit become available.]
(Hmmm, this is only in the PDF-version ...). May be there is no point
in providing this executable at all - at least concentrating for the
tests on other binaries makes sense.
The third test in your run-unit-test script is simply calling cd-hit-div
with wrong parameters. If you look into the very simple Perl code it
does not use any options like -i or -o but rather takes the arguments
as input output and div number. So
cd-hit-div example.fa output 4
creates some result (even if I have no idea whether it makes sense.
In short:
1. The current executable name is cdhit and there are other
executables without the '-'. I'm tempted to change this
to fit the upstream names. Please checkout if you can
write more tests for these.
2. I have no idea why the cd-hit-para tests are failing. If
there is no answer here please contact upstream
3. Please verify that my fix for the cd-hit-div call makes
sense.
Kind regards
Andreas.
--
http://fam-tille.de
Reply to: