Age | Commit message (Collapse) | Author |
|
|
|
Fixes: https://gitlab.desy.de/thomas.white/crystfel/-/issues/103
|
|
Previously, it counted the number of "COMPLETED" jobs. This is
unreliable, because sub-jobs might be taken out of the scontrol output
before the whole job is complete. Now, it counts the PENDING (and
similar) states, and subtracts from the total number of tasks to get a
reliable count.
Fixes: https://gitlab.desy.de/thomas.white/crystfel/-/issues/102
|
|
This brings the GUI and manual pages back into alignment (so to speak)
with the changes in d00a9f8566bd1dd49d88cffe3580dfa8d4b27299.
align_detector --out-of-plane-tilts is not expected to be very stable,
and currently isn't available via the GUI.
|
|
This was broken by 1b8abebf8bf37d5d57ed55816223d95557b7f844.
|
|
|
|
|
|
This slightly reduces spam in Valgrind output.
|
|
On my system, the icons were showing up too small, but technically at
the correct size (48 pixels). Now, they can be drawn at whatever scale
fits the screen best, as judged by GTK. For me, this restores the
correct size.
|
|
Otherwise, we run the risk of a very wide window.
|
|
It makes the display a bit too crowded, and it seems less likely that
the indices will be needed than not.
|
|
This changes the behaviour of --out-of-plane to refine only the panel
translations in the z-direction, not rotations.
Refining panel tilts out-of-plane seems to be a bit of a stretch, but
there is a new option --out-of-plane-tilts to try it.
|
|
This extends the "Jump to frame" dialogue box by adding two things:
1. A way to choose a file, if there is more than one.
Only the first 20 unique filenames are considered, to avoid a very
long list.
2. A way to choose a frame number (as shown in the filename bar), rather
than a frame ID.
Closes: https://gitlab.desy.de/thomas.white/crystfel/-/issues/104
|
|
Fixes: https://gitlab.desy.de/thomas.white/crystfel/-/issues/106
|
|
It's needed for detector refinement.
Fixes: https://gitlab.desy.de/thomas.white/crystfel/-/issues/101
|
|
I was noticing a segmentation fault appearing at the end of an
indexamajig call. Debugging gave me:
snprintf (__fmt=0x40d1fd "%s/mosflm.lp", __n=127, __s=0x471f90 "")
delete_temporary_folder (n_proc=12, tmpdir=0x427c00 "./indexamajig.260408") at ../src/im-sandbox.c:1100
Which led me to these snprintf statements. Easily fixed, but a bit had
to spot.
|
|
There was a truly ancient bug here, which has existed for as long as
CrystFEL has supported multi-crystal indexing (version 0.5.0, 11 years
ago!). See 1e03ed982741fdc576ec5a915da120450df20499.
Previously, when using process_hkl --even-only (without --start-after),
if the first image contained multiple crystals, they would all be
skipped. The correct behaviour is for second, fourth, sixth etc
crystals to be merged, if they exist.
If the first image contained exactly two crystals, this glitch would put
the alternating sequence one step out of sync, such that the only
difference between --odd-only and --even-only was the inclusion of the
very first crystal. Fortunately, it was obvious when this happened,
because the two merged files would be practically identical.
|
|
|
|
|
|
Prevously, it disabled the resend timeout, but still sent the
acknowledgements.
|
|
|
|
With thanks to biochem-fan.
Fixes: https://github.com/taw10/crystfel/issues/12
|
|
|
|
Otherwise, we end up acknowledging the previous message ID after
receiving something like kEndOfStream.
|
|
Something strange is going on here, need to debug.
|
|
|
|
Fixes: https://github.com/taw10/crystfel/issues/8
|
|
|
|
|
|
|
|
|
|
|
|
With large indexing jobs, we generate a LOT (4000+) of Millepede files.
This option is needed to prevent pede from keeping all these files open
simultaneously, which runs into an operating system limit.
|
|
Depending on the backend, this could be a single "mille-data", or
multiple "mille-data-X" folders.
|
|
Previously, they all wrote into the same "mille-data" folder,
overwriting each others' files and causing general carnage.
|
|
This changes all the peak search procedures into pure functions that
return a new ImageFeatureList. This takes the management of
image->features out of the hands of the peak search routines, and into
the calling code's responsibility.
In turn, this allows a load of stuff to become const.
|
|
|
|
This also gets rid of crystal_copy_deep. From now on, all crystal_copy
calls also copy the UnitCell.
|
|
This fixes the remaining programs.
|
|
This fixes the entire partialator/scaling/rejection part.
|
|
This fixes integration.c, peaks.c and process_hkl.c.
|
|
|
|
|
|
|
|
|
|
Indexamajig --help causes a segfault on MacOS due to a bug in
argp-standalone. That package is unlikely to be updated any time soon.
Eventually, Homebrew might switch to one of the non-crashy forks, but
probably not soon. We aren't the only ones affected - see
https://www.openwall.com/lists/musl/2021/02/10/1
This commit adds a warning message to state that we already know about
it, and it's not "our" fault. It also removes the corresponding test,
since we know it fails.
|
|
|
|
|
|
|
|
|