Bug #1094
make check from gammalib-00-08-00 has one failure
| Status: | Closed | Start date: | 01/21/2014 | |
|---|---|---|---|---|
| Priority: | Normal | Due date: | ||
| Assigned To: |  Knödlseder Jürgen | % Done: | 100% | |
| Category: | - | |||
| Target version: | 00-08-01 | |||
| Duration: |
Description
Test binned observation: ..E. NOK
Recurrence
No recurrence.
Related issues
History
#1
Updated by Knödlseder Jürgen almost 11 years ago
- Project changed from ctools to GammaLib
#2
Updated by Knödlseder Jürgen almost 11 years ago
Johann, could you please upload the files test/test_COM.log and test/reports/GCOM.xml so that I can check what’s going on? It seems that Stefan Klepser has encountered the same problem. I’m wondering why I don’t see this on my machine ...
#4
Updated by Cohen-Tanugi Johann almost 11 years ago
- File test_COM added
#5
Updated by Knödlseder Jürgen almost 11 years ago
- File config.log added
For the record, I grabbed the gammalib-00-08-00 release tarball and build it on my Mac OS X 10.6.8. All tests passed. Here the configuration file: config.log
#6
Updated by Cohen-Tanugi Johann almost 11 years ago
- File config.log added
ok here is mine. The first thing is that we do not use the same gcc version, and I am on 64bits while you are on 32bits. Beyond that, I see that there are compilation failures in the logs, for both of us.....
#7
Updated by Knödlseder Jürgen almost 11 years ago
Okay, problem is as I expected in Test XML constructor:
<testcase assertions="" classname="COMPTEL instrument specific class testing" name="Test binned observation: Test XML constructor" status="" time="0">
24
<error message='*** ERROR in GFits::open(std::string&): Unable to open FITS file "/home/cohen/sources/CTA/CTOOLS/ctools-00-07-00/share/caldb/u47569_iaq.fits" (status=104)' type="N10GException15fits_open_errorE" />
25
</testcase>
It tries to open the file
/home/cohen/sources/CTA/CTOOLS/ctools-00-07-00/share/caldb/u47569_iaq.fits. In the XML file, the filename is u47569_iaq.fits. The code apparently adds the path /home/cohen/sources/CTA/CTOOLS/ctools-00-07-00/share/caldb which is, I guess, the GAMMALIB_CALDB environment variable. Here the relevant code
void GCOMResponse::load(const std::string& iaqname)
{
// Save calibration database name
std::string caldb = m_caldb;
// Clear instance
clear();
// Restore calibration database name
m_caldb = caldb;
// Save IAQ name
m_iaqname = iaqname;
// Build filename
std::string filename = m_caldb + "/" + m_iaqname;
// Open FITS file
GFits file(filename);
// Get IAQ image
const GFitsImage& iaq = *file.image(0);
// Read IAQ
read(iaq);
// Close ARF FITS file
file.close();
// Return
return;
}
m_caldb is set by the std::string GCaldb::rootdir() method and is either the environment variable GAMMALIB_CALDB or CALDB (whichever exists in the given order). The GAMMALIB_CALDB is set by the GammaLib configuration script. CALDB is set by the ctools configuration script. Maybe this is the problem?#8
Updated by Knödlseder Jürgen almost 11 years ago
- Assigned To set to Knödlseder Jürgen
- % Done changed from 0 to 10
I checked GCaldb::rootdir() again and noticed that in contrary to expectations, CALDB takes precedence over GAMMALIB_CALDB. I still need to dive more in the code to understand the problem. For now I try to reproduce this by setting the CALDB environment variable.
#9
Updated by Knödlseder Jürgen almost 11 years ago
Got it. Setting the CALDB environment variable gave
********************************************* * COMPTEL instrument specific class testing * ********************************************* Test instrument direction: ............ ok Test response: ... ok Test binned observation: ..E. NOK Test event bin: ............ ok Test event cube: .................... ok
#10
Updated by Knödlseder Jürgen almost 11 years ago
- Target version set to 00-08-01
- % Done changed from 10 to 20
And understood. The COMPTEL unit test sets internally the GAMMALIB_CALDB environment variable before jumping into the code:
// Set GAMMALIB_CALDB environment variable
std::string caldb = "GAMMALIB_CALDB="+com_caldb;
putenv((char*)caldb.c_str());
This should work if
GAMMALIB_CALDB takes precedence over CALDB. As the code is in fact not doing this, CALDB is used if it’s set, directing to a directory that does not contain the response files.
So the immediate correction is to correct the GCaldb::rootdir() method.
Follow-up actions should be taken to clean up the CALDB interface. Actually, the expected structure
$CALDB/data/<mission> $CALDB/data/<mission>/<instrument> $GAMMALIB_CALDB/data/<mission> $GAMMALIB_CALDB/data/<mission>/<instrument>is not really used (the
data directory is actually missing), and the GCOMResponse response class does not use the appropriate GCaldb::path() method to access the data.
The latter is followed up by change request #1102.
#11
Updated by Knödlseder Jürgen almost 11 years ago
- Status changed from New to Closed
#12
Updated by Knödlseder Jürgen almost 11 years ago
- % Done changed from 20 to 100