[med-svn] r23642 - trunk/packages/metastudent/trunk/debian

[Tatiana Malygina]

Author: latticetower-guest
Date: 2017-01-19 11:17:56 +0000 (Thu, 19 Jan 2017)
New Revision: 23642

Modified:
   trunk/packages/metastudent/trunk/debian/README.Debian
   trunk/packages/metastudent/trunk/debian/README.test
   trunk/packages/metastudent/trunk/debian/changelog
Log:
add details on package testsuite failure to README.Debian, update README.test

Modified: trunk/packages/metastudent/trunk/debian/README.Debian
===================================================================
--- trunk/packages/metastudent/trunk/debian/README.Debian	2017-01-18 16:34:53 UTC (rev 23641)
+++ trunk/packages/metastudent/trunk/debian/README.Debian	2017-01-19 11:17:56 UTC (rev 23642)
@@ -3,3 +3,47 @@
 metastudent searches through large data files (>200M compressed) in order to predict gene ontology terms. These data files are available from the metastudent-data package or [1], and are released under the GPL-2+.
 
 [1] ftp://rostlab.org/metastudent/metastudent-data_[0-9.]+.tar.gz
+
+
+Metastudent currently works correctly only with legacy version of blast. It is
+ stated in metastudent development repository here: 
+
+    https://github.com/Rostlab/MetaStudent
+
+Since the moment when blast2 package is resolved as blast+ set of programs,
+autopkgtest testsuite began to fail. The main reason is that blast+ programs
+produce slightly different output. Particularly, in some lines of these files
+number of space characters differs from those found in legacy blast output,
+this leads to the main problem that regular expressions, used by metastudent 
+to parse these lines, don't parse them correctly. This leads to incorrect
+metastudent output. 
+ 
+Another problem, which is the reason why Bug #848903 appeared, is also related
+to difference in programs output. When legacy versions of blast produce output
+for several input .fasta sequences, at the beginning of each sequence's input
+set they contain line starting with "BLASTP" substring (that's a program name
+used by legacy blast). Metastudent searches for this substring and considers
+data following this line to be related to next protein sequence in input .fasta
+file. This is relatively simplier to fix, since it produces only 1 exception
+which is the reason of Bug #848903. 
+
+Problem with blast output parsing appears in several places of metastudent,
+since different code is used to parsing in "studentA" and "studentB" parts.
+
+Currently there is also a problem with test files, since the result files
+provided by upstream and named "test.result.BPO.txt", "test.result.CCO.txt"
+and "test.result.MFO.txt" - contain result of metastudent run only for
+1 sequence (instead of full output for 3 sequences given in "test.fasta").
+
+Necessary steps to fix everything to produce correct results with new version
+of blast2 package (work in progress):
+
+1. Add patch for fixing bug #848903
+2. Run metastudent with legacy blast 2.2.26 and save all results, including
+temporary, for latest version of metastudent-data package. They should be
+similar to expected results produced by blast+ on the same version of
+metastudent-data package and might be used as a reference while fixing package.
+3. Metastudent runs in 3 stages, and they use different code for parsing blast
+run output. Check them one by one and fix regular expressions, at each stage
+compare temporary results with those produced by metastudent run with
+legacy blast.

Modified: trunk/packages/metastudent/trunk/debian/README.test
===================================================================
--- trunk/packages/metastudent/trunk/debian/README.test	2017-01-18 16:34:53 UTC (rev 23641)
+++ trunk/packages/metastudent/trunk/debian/README.test	2017-01-19 11:17:56 UTC (rev 23642)
@@ -7,7 +7,24 @@
 
 in order to confirm its integrity.
 
-To test manually, cd to some dir where you have write permissions and do
+Currently metastudent supports only legacy version of blast (for additional
+details see README.Debian). That's why currently autopkgtest fails and this
+package should be tested and run manually.
+
+There are several necessary steps to run it now, if you expect correct results:
+
+1. Install legacy blast from NCBI website. 
+  I recommend using legacy blast 2.2.26, since I checked metastudent results
+  with this specific version).
+2. Copy metastudentrc.default config file from metastudent dir to location 
+  where you have write permissions. 
+3. Find "blastpgp" file in legacy blast installation. Edit your 
+  metastudentrc.default copy to point to this file in  line starting with
+  "BLASTPGP_EXE_PATH".
+
+After that, do
  
     cp /usr/share/doc/metastudent/examples/test.fasta .
-    metastudent -i test.fasta -o test.result --temp-dir=.
+    metastudent -i test.fasta -o test.result --temp-dir=. --config=metastudentrc.default
+
+The result files produced by metastudent should appear in working dir.

Modified: trunk/packages/metastudent/trunk/debian/changelog
===================================================================
--- trunk/packages/metastudent/trunk/debian/changelog	2017-01-18 16:34:53 UTC (rev 23641)
+++ trunk/packages/metastudent/trunk/debian/changelog	2017-01-19 11:17:56 UTC (rev 23642)
@@ -2,6 +2,9 @@
 
   * Team upload.
   * add allow-stderr restriction to debian/tests/control
+  * add details on package testsuite failure to REAMDE.Debian,
+    add to README.test how to test manually (currently metastudent supports
+    legacy blast only and fails on blast+ output parsing).
 
  -- Tatiana Malygina <merlettaia at gmail.com>  Sat, 16 Jul 2016 09:24:57 +0300