[IGSMAIL-5615]: A harmful bug in CRX2RNX was found and fixed
Yuki Hatanaka
hata at gsi.go.jp
Fri Jun 22 03:14:18 PDT 2007
******************************************************************************
IGS Electronic Mail 22 Jun 03:14:29 PDT 2007 Message Number 5615
******************************************************************************
Author: Yuki Hatanaka
Dear colleagues,
A harmful bug was found in the RINEX uncompression program
CRX2RNX. The bug resides only in the crx2rnx.c versions
4.0.0-4.0.2. Since rnx2crx.c is free from this bug, the
Compact RINEX files just compressed with RNX2CRX of
these versions are not affected (unless re-compressed
after uncompression).
I'm sorry for the trouble the bug may cause.
All the package files in the GSI's FTP server was replaced
with bug fixed ones with updating the version of software
from 4.0.2 to 4.0.3:
ftp://terras.gsi.go.jp/software/RNXCMP_4.0.3/
The followings are the detailes about the bug.
Reported problem:
CRX2RNX version 4.0.2 produced corrupted RINEX files
when being applied to some of Compact RINEX files of
1 Hz sampling, although an old version (2.4.3) of
CRX2RNX uncompressed the same file correctly.
The cause:
In process of uncompress of a data arc, value of some
specific intermediate variables may drift in a
randome-walk-like manner and may overflow eventually
and corrupt the recovered output of RINEX files.
The drift was caused by omission of carrying numbers
in a intermediate process. The higher the sampling rate
is, the more likely an overflow may take place, hence.
Solution:
Drift and overflow of the intermediate variable was
prevented by adding process of carrying numbers.
new version 4.0.3 is created.
Finally, I would thank Mark Schenewerk for reporting
the problem and providing a sample file, which was
helpful to identify the cause of the bug.
Best Regards,
Yuki
----------------------------------------------------
Yuki Hatanaka, Ph.D, Geographical Survey Institute
e-mail : hata at gsi.go.jp
---------------------------------------------------
More information about the IGSMail
mailing list