The bugs mostly involved using the config apparatus to generate multiple output files, especially when the number of objects per file is variable. We now do this deliberately in demo9 to help test that this kind of action is done correctly.
Changes from v1.0.0 to v1.0.1:
--------------------------
* Fixed some bugs in the config machinery when files have varying numbers
of objects. (#487)
- If the number of objects is dependent on a Dict value, the code had been
erroneously using the Dict for the previous file.
- Sequences that index on the object number had not necessarily been
starting at the first object number in a file when the files had varying
numbers of objects.
- There had not been a random number generator available in the config
for items at the file-level scope. So if you wanted nobjects to be
a random variate, that had not been possible.
* Fixed a bug in config where objects that are considered "safe" (that is,
unchanging from object to object -- the psf profile, for example) during
the processing of a given file were not being correctly invalidated for
the next file in a multi-file run if the next file uses a different
catalog for instance.
* Support astropy.io.fits in lieu of stand-alone pyfits module. (#488)
This is where pyfits will live going forward. So we now support both
the astropy distribution as well as the legacy pyfits module.
* Fixed a bug in the drawing of a Pixel all by itself. It had been rendered
incorrectly, although the bug did not affect the rendering of other profiles
convolved by a Pixel. (#497)