Brian Savage
All,
A few more bugs were found in sac. They have been fixed and should be
available in the next release. They are not in the repository yet, but
will be soon.
All of these bug are short, quick fixes. Please email me and I will
send you the appropriate files.
Bug List: ( explanations below )
- Memory Leak Bug
- Power of 2 Bug
- read more Bug
Cheers,
Brian
Memory Leak Bug:
-------------------------
Multiple read commands exposes a memory leak which can become very
large, very quickly. I found this bug by unwittingly consuming 4 GB of
memory while processing a large dataset.
Example:
SAC> read file1
SAC> read file2
...
SAC> read filen
Fix:
The allocated memory for storing the sac data was not freed when
executing another read call, not read more. Calls to relamb were
included which releases memory or frees it back to the system. this was
done in dfm/readfl().
Power of 2 Bug:
-----------------------
Determining the next largest power of 2 for numbers slightly larger than
a power of 2 returns the smaller power of 2. The bug is in ucf/next2().
It can cause memory error and buffer overruns, causing sac to crash or
experience unexpected behavior.
Example:
Number of points: 8193
Get next largest power of 2: (should be 16384)
ucf/next2() returns 8192
Fix:
The cause of the bug was a rounding error of a real number to an
integer. A call to ceil() fixes the bug, plus a check to make certain
the returned power of 2 is larger than the initial number of points.
read more Bug:
-----------------------
While reading a large number of files into sac using the "read more",
sac grinds to a halt. The problem was found in the SeisMgr code as it
was looping over the list of files during each read more command.
Example:
SAC> read more file1
SAC> read more file2
...
SAC> read more filen
Fix:
No fix for this bug has been found, but turning off the SeisMgr code
essentially fixes the problem and does not harm to the working of sac.
My working copy has an environment variable to indicate whether or not
to use the SeisMgr code. I am not certain what the advantages the
SeisMgr code provides the sac code. If someone has a list of the
benefits that would be great. The recall bug, still not fixed, is due
to the same section of code.
A few more bugs were found in sac. They have been fixed and should be
available in the next release. They are not in the repository yet, but
will be soon.
All of these bug are short, quick fixes. Please email me and I will
send you the appropriate files.
Bug List: ( explanations below )
- Memory Leak Bug
- Power of 2 Bug
- read more Bug
Cheers,
Brian
Memory Leak Bug:
-------------------------
Multiple read commands exposes a memory leak which can become very
large, very quickly. I found this bug by unwittingly consuming 4 GB of
memory while processing a large dataset.
Example:
SAC> read file1
SAC> read file2
...
SAC> read filen
Fix:
The allocated memory for storing the sac data was not freed when
executing another read call, not read more. Calls to relamb were
included which releases memory or frees it back to the system. this was
done in dfm/readfl().
Power of 2 Bug:
-----------------------
Determining the next largest power of 2 for numbers slightly larger than
a power of 2 returns the smaller power of 2. The bug is in ucf/next2().
It can cause memory error and buffer overruns, causing sac to crash or
experience unexpected behavior.
Example:
Number of points: 8193
Get next largest power of 2: (should be 16384)
ucf/next2() returns 8192
Fix:
The cause of the bug was a rounding error of a real number to an
integer. A call to ceil() fixes the bug, plus a check to make certain
the returned power of 2 is larger than the initial number of points.
read more Bug:
-----------------------
While reading a large number of files into sac using the "read more",
sac grinds to a halt. The problem was found in the SeisMgr code as it
was looping over the list of files during each read more command.
Example:
SAC> read more file1
SAC> read more file2
...
SAC> read more filen
Fix:
No fix for this bug has been found, but turning off the SeisMgr code
essentially fixes the problem and does not harm to the working of sac.
My working copy has an environment variable to indicate whether or not
to use the SeisMgr code. I am not certain what the advantages the
SeisMgr code provides the sac code. If someone has a list of the
benefits that would be great. The recall bug, still not fixed, is due
to the same section of code.
-
Brian,
Nice work fixing these bugs.
FYI, the SeisMgr code is essential when using a relational database
as is done by the GNEM program at LLNL.
While it isn't something that is currently useful to the IRIS
community, it seems to me that it would be
good to try and maintain this capability at some level so that it can
be utilized should it be needed.
Given the growing DMC archive and the amount of data coming available
from USArray I would think
that this is a realistic possibility.
Using an environment variable to turn it on or off seems like a good
way to keep it in the code without
impacting those who don't use it.
Cheers,
Peter
At 11:19 AM -0400 6/30/06, Brian Savage wrote:
All,
--
A few more bugs were found in sac. They have been fixed and should be
available in the next release. They are not in the repository yet, but
will be soon.
All of these bug are short, quick fixes. Please email me and I will
send you the appropriate files.
Bug List: ( explanations below )
- Memory Leak Bug
- Power of 2 Bug
- read more Bug
Cheers,
Brian
Memory Leak Bug:
-------------------------
Multiple read commands exposes a memory leak which can become very
large, very quickly. I found this bug by unwittingly consuming 4 GB of
memory while processing a large dataset.
Example:
SAC> read file1
SAC> read file2
...
SAC> read filen
Fix:
The allocated memory for storing the sac data was not freed when
executing another read call, not read more. Calls to relamb were
included which releases memory or frees it back to the system. this was
done in dfm/readfl().
Power of 2 Bug:
-----------------------
Determining the next largest power of 2 for numbers slightly larger than
a power of 2 returns the smaller power of 2. The bug is in ucf/next2().
It can cause memory error and buffer overruns, causing sac to crash or
experience unexpected behavior.
Example:
Number of points: 8193
Get next largest power of 2: (should be 16384)
ucf/next2() returns 8192
Fix:
The cause of the bug was a rounding error of a real number to an
integer. A call to ceil() fixes the bug, plus a check to make certain
the returned power of 2 is larger than the initial number of points.
read more Bug:
-----------------------
While reading a large number of files into sac using the "read more",
sac grinds to a halt. The problem was found in the SeisMgr code as it
was looping over the list of files during each read more command.
Example:
SAC> read more file1
SAC> read more file2
...
SAC> read more filen
Fix:
No fix for this bug has been found, but turning off the SeisMgr code
essentially fixes the problem and does not harm to the working of sac.
My working copy has an environment variable to indicate whether or not
to use the SeisMgr code. I am not certain what the advantages the
SeisMgr code provides the sac code. If someone has a list of the
benefits that would be great. The recall bug, still not fixed, is due
to the same section of code.
_______________________________________________
sac-dev mailing list
sac-dev<at>iris.washington.edu
http://www.iris.washington.edu/mailman/listinfo/sac-dev
Peter Goldstein, Ph.D. (925) 423-1231 (office)
L-103, PO Box 808 (925) 422-5844 (fax)
Livermore, CA 94551 peterg<at>llnl.gov (email)
web page: http://www.llnl.gov/sac
Peter Goldstein
