[Mnj] CVEL bugs
Eric Greisen
egreisen at nrao.edu
Mon Mar 1 18:17:05 EST 2010
CVEL has been found to have a bug whose result is obvious and one which
was more serious and not obvious.
1. The obvious bug, introduced in 2009, caused the task to go into an
infinite loop while handling FO table data. It was caused by the
inaccuracies of computing on times and was easily corrected by adding a
fudge factor and other checks.
2. The non-obvious bug caused the task to do no shifting for scans after
the first in single-source data sets with an index (NX) table. These
tables have been present on many/most single-source files for the past
1-3 years. The failure was obscured by a minor error in which the
source name was not set. The resulting null characters truncated the
message that should have said "not shifted". This is a serious error
and all use of CVEL on single-source files must be repeated.
The patch to 31DEC09 is available now, either with local compilation or
through the MNJ. 31DEC10 will be fixed tonight and a MNJ tomorrow will
correct your code.
Eric Greisen
More information about the Mnj
mailing list