SANS: Crash - bad_alloc in LoadSampleDetailsFromRaw

[Russell Taylor]

From Steve King:

Back end of log output on batch run......

Algorithm: LoadRaw v3

Parameters:

Name: Filename, Value: N:/cycle_09_4\SANS2D00002321.raw, Default?: No, Direction: Input

Name: OutputWorkspace, Value: 2321_sans_raw, Default?: No, Direction: Output

Name: SpectrumMin, Value: 1, Default?: Yes, Direction: Input

Name: SpectrumMax, Value: 2147483632, Default?: Yes, Direction: Input

Name: SpectrumList, Value: , Default?: Yes, Direction: Input

Name: Cache, Value: If Slow, Default?: Yes, Direction: Input

Name: LoadLogFiles, Value: 1, Default?: Yes, Direction: Input

Name: LoadMonitors, Value: Include, Default?: Yes, Direction: Input

* Run title: 02321 Richardson_{Gavin Empty Cell 3.5m_SANS 26-NOV-2009 21:42:24 *}

Cannot open ICPevent file N:/SANS2D00002321_ICPevent.txt. Period 1 assumed for all data.

Unable to open file N:/SANS2D00002321_Changer.txt

Unable to open file N:/SANS2D00002321_Fast_Shutter.txt

Unable to open file N:/SANS2D00002321_Front_Det_Temp.txt

Unable to open file N:/SANS2D00002321_Height.txt

Unable to open file N:/SANS2D00002321_ICPdebug.txt

Unable to open file N:/SANS2D00002321_ICPevent.txt

Unable to open file N:/SANS2D00002321_ICPstatus.txt

Unable to open file N:/SANS2D00002321_Julabo.txt

Unable to open file N:/SANS2D00002321_Moderator_Temp.txt

Unable to open file N:/SANS2D00002321_Rear_Det_Temp.txt

Unable to open file N:/SANS2D00002321_Sample.txt

Unable to open file N:/SANS2D00002321_Status.txt

Unable to open file N:/SANS2D00002321_Table.txt

Data Object '2321_sans_raw' replaced in data service.

LoadRaw successful, Duration 0.797 seconds

Algorithm: LoadSampleDetailsFromRaw v1

Parameters:

Name: InputWorkspace, Value: 2321_sans_raw, Default?: No, Direction: Input

Name: Filename, Value: N:/cycle_09_4\SANS2D00002321.raw, Default?: No, Direction: Input

Error in execution of algorithm LoadSampleDetailsFromRaw:

bad allocation

[Russell Taylor]

(In [4217]) Couple more places where files weren't being closed. Re #1177.

[Russell Taylor]

Hopefully that was the problem, but I haven't tested it at all (and certainly not in batch mode).

[Martyn Gigg]

Bad alloc now occurs on running LoadRaw. Simple test script (providing the SANS2D data directory in the repository is in the Mantid.properties path)

run_number = 992
prefix = "SANS2D00000"
ext = ".raw"
output_name = "sans"
for i in range(0,15):
    LoadRaw(prefix + str(run_number) + ext, output_name + "_" + str(i))

After about 9/10 the bad alloc should occur

[Russell Taylor]

What's happening in the specific test above, where the loop loads into a different workspace each time, is that we really are running out of memory.

At first workspaces go into memory. After a while they start going 'ManagedRaw', but even in this case some memory is used (e.g. the spectra-detector map) so the footprint continues to rise, albeit much slower, and eventually a bad_alloc arises. In this case, the problem could be mitigated by lowering the threshold for managed workspaces in the properties file.

[Stuart Campbell]

This ticket has been transferred to github issue ​2024