taw10 / crystfel Goto Github PK

When I press "settings" button in the freshly oppened GUI it crashes with "segmentation fault" error.

In align_detector, you are dereferencing %s without a variable.

Even with --out-of-plane, refined coffset is not written to the output geometry file.

The cause is this check:

If groups are defined before its panels are defined, we get this error:

Unknown panel group 'q1'
Unknown panel group 'q2'
...
Invalid top-level line 'group_all'

Rigid body definitions did not have this order requirement.

Is this by design? If so, it would be nice to document this.

In our SACLA MPCCD geometry file, we have this for the second panel (q2):

data = /%/data

q2/adu_per_eV = 0.000999
q2/min_fs = 0
q2/min_ss = 1024
q2/max_fs = 511
q2/max_ss = 2047
q2/fs = 0.006215x -0.999981y
q2/ss = 0.999981x +0.006215y
q2/corner_x = 36.444600
q2/corner_y = 497.856016

badq2l1/min_fs = 0
badq2l1/max_fs = 4
badq2l1/min_ss = 1024
badq2l1/max_ss = 2047
badq2l1/panel  = q2

Note that eight 512 x 1024 px panels are stacked along the slow (1024 px) axis into a HDF5 dataset.

align_detector rewrites this bad region definition as follows:

bad_badq2l1/panel = q2
bad_badq2l1/min_fs = 0
bad_badq2l1/max_fs = 4
bad_badq2l1/min_ss = 0
bad_badq2l1/max_ss = 1023

It seems that min_ss and max_ss have been rewritten in terms of the panel's local data coordinate.

However, indexamajig still interprets them in terms of the global data coordinate system.

Bad region 'bad_badq2l1' is outside the panel bounds (q2) as presented in data (0 1024, 511 2047 inclusive): Bad region 0,0 to 4, 1023 inclusive

I don't know which is intended. Earlier versions worked in the same way as indexamajig, which I prefer.

indexamajig --input=i to read file lists from STDIN is broken. We got Couldn't open '-'.

The following patch will fix this problem.

index aa0db21..0dfc295 100644
--- a/src/indexamajig.c
+++ b/src/indexamajig.c
@@ -1241,7 +1241,7 @@ int main(int argc, char *argv[])
                return 1;
        }
 
-       if ( (args.filename != NULL) && is_hdf5_file(args.filename, &err) ) {
+       if ( (args.filename != NULL) && (strcmp(args.filename, "-") != 0) && is_hdf5_file(args.filename, &err) ) {
                ERROR("Your input file appears to be an HDF5 file.\n");
                ERROR("The input file should be a list of data files, not the "
                      "data file itself.\n");

It would be great if one can fill the path to data by copy-paste instead of navigating through folders.

I was working with SLAC/mfxp17318_0015 data. After all indexing with determined unit cell file, I got Subprocess exited with status 512 error with no message while running Indexing Ambiguity worker.

I couldn't debug it yet for the reason of error, but, there should be a log of exception.

Is there any known possible reason of the error?

When there is no detector hierarchy information in a geometry file, indexamajig --mille rightly complains:

Cannot write calibration data (--mille) because geometry does not contain hierarchy information (group 'all' not found)

However, non zero-byte mille files are actually written. Passing this to align-detector causes a segmentation fault.

If the detector consists of only one panel, CrystFEL will automatically create the all group containing it.
https://gitlab.desy.de/thomas.white/crystfel/-/blob/master/doc/man/crystfel_geometry.5.md

How about assuming all panels belong to the all group if there is no definition?