After discovering that one of the faces in one of the host/user directories
had it's left and right sides inverted, I began to wonder how many other
faces I'd incorrectly converted from 48x48x1 format, to face.xbm format.
For those of you familiar with Open Windows, you'll know there is a DeskSet
utility called filemgr. With the V2 filemgr, there is a "view by content"
option for looking at the contents of monochrome rasterfiles as shrunken
64x64 icons.
I've extended V3 filemgr to support other formats; namely Sun icon format,
XBM format and XPM format. The XBM "view by content" support came in very
handy for checking which files were not displaying correctly.
I sampled one face.xbm file for one user for each host. I made the
assumption that if one users face.xbm file was incorrect, probably every
face.xbm file for that host was incorrect. Not a perfect assumption but
reasonable.
I then created some shell scripts to convert the broken face.xbm files back
to 48x48x1 files, then reconvert back to face.xbm, this time correcting the
endian problem.
There were 653 files converted incorrectly in all. I'd like to think the
database is in a *lot* better shape now.
I've re-put the database as a compressed tar file on
titan:incoming/facedir.tar.Z
and iuvax.cs.indiana.edu:pub/faces/incoming/facedir.tar.Z
There will still be small teething problems. If you find any, please let
me know and I'll adjust accordingly.
Now I can rest easier.