smooth() bug in (at least) version 8.5.1 [message #93939] |
Fri, 02 December 2016 06:03  |
Helder Marchetto
Messages: 520 Registered: November 2011
|
Senior Member |
|
|
Hi,
I came across this bug and it's a "heavy" one, meaning that IDL crashes and you're thrown back out.
The crashing commands are (try at own risk!):
sm = smooth(indgen(512,511), 9, /edge_mirror)
sm = smooth(indgen(512,510), 9, /edge_mirror)
And just to be pedantic, I've tested a few other cases and these did NOT result in a crash:
sm = smooth(indgen(512,512), 9, /edge_mirror)
sm = smooth(indgen(511,511), 9, /edge_mirror)
It appears that smooth (stopped?) being able to smooth rectangular arrays.
By running the above from the command line I get the following error (for a couple of seconds on the command line terminal):
% Array has a corrupted descriptor: <no name>
% Execution halted at: $Main$
And my IDL version.
IDL> !version
{
"ARCH": "x86_64",
"OS": "Win32",
"OS_FAMILY": "Windows",
"OS_NAME": "Microsoft Windows",
"RELEASE": "8.5.1",
"BUILD_DATE": "Nov 14 2015",
"MEMORY_BITS": 64,
"FILE_OFFSET_BITS": 64
}
I others are brave enough, could you let me know if it crashes on other versions of IDL? Edge_mirror and _wrap have been added only in 8.1.
Cheers,
Helder
|
|
|
Re: smooth() bug in (at least) version 8.5.1 [message #93941 is a reply to message #93939] |
Fri, 02 December 2016 07:06   |
wlandsman
Messages: 743 Registered: June 2000
|
Senior Member |
|
|
On Friday, December 2, 2016 at 9:03:27 AM UTC-5, Helder wrote:
> Hi,
> I came across this bug and it's a "heavy" one, meaning that IDL crashes and you're thrown back out.
>
> The crashing commands are (try at own risk!):
> sm = smooth(indgen(512,511), 9, /edge_mirror)
> sm = smooth(indgen(512,510), 9, /edge_mirror)
>
> And just to be pedantic, I've tested a few other cases and these did NOT result in a crash:
> sm = smooth(indgen(512,512), 9, /edge_mirror)
> sm = smooth(indgen(511,511), 9, /edge_mirror)
>
> It appears that smooth (stopped?) being able to smooth rectangular arrays.
>
> By running the above from the command line I get the following error (for a couple of seconds on the command line terminal):
> % Array has a corrupted descriptor: <no name>
> % Execution halted at: $Main$
>
> And my IDL version.
> IDL> !version
> {
> "ARCH": "x86_64",
> "OS": "Win32",
> "OS_FAMILY": "Windows",
> "OS_NAME": "Microsoft Windows",
> "RELEASE": "8.5.1",
> "BUILD_DATE": "Nov 14 2015",
> "MEMORY_BITS": 64,
> "FILE_OFFSET_BITS": 64
> }
>
> I others are brave enough, could you let me know if it crashes on other versions of IDL? Edge_mirror and _wrap have been added only in 8.1.
>
> Cheers,
> Helder
It also crashes on my Macbook running 8.5.1 and on my Linux systems running (32 bit) 8.2.2 and (64 bit) 8.5.1. So it looks like a problem in the internal SMOOTH() algorithm. --Wayne
"ARCH": "x86_64",
"OS": "darwin",
"OS_FAMILY": "unix",
"OS_NAME": "Mac OS X",
"RELEASE": "8.5.1",
"BUILD_DATE": "Nov 14 2015",
"MEMORY_BITS": 64,
"FILE_OFFSET_BITS": 64
ARCH STRinG 'x86_64'
OS STRING 'linux'
OS_FAMILY STRING 'unix'
OS_NAME STRING 'linux'
RELEASE STRING '8.5.1'
BUILD_DATE STRING 'Nov 14 2015'
MEMORY_BITS INT 64
FILE_OFFSET_BITS
INT 64
ARCH STRING 'x86'
OS STRING 'linux'
OS_FAMILY STRING 'unix'
OS_NAME STRING 'linux'
RELEASE STRING '8.2.2'
BUILD_DATE STRING 'Jan 23 2013'
MEMORY_BITS INT 32
FILE_OFFSET_BITS
INT 64
|
|
|
Re: smooth() bug in (at least) version 8.5.1 [message #93942 is a reply to message #93941] |
Fri, 02 December 2016 07:13   |
Helder Marchetto
Messages: 520 Registered: November 2011
|
Senior Member |
|
|
On Friday, December 2, 2016 at 4:06:08 PM UTC+1, wlandsman wrote:
> On Friday, December 2, 2016 at 9:03:27 AM UTC-5, Helder wrote:
>> Hi,
>> I came across this bug and it's a "heavy" one, meaning that IDL crashes and you're thrown back out.
>>
>> The crashing commands are (try at own risk!):
>> sm = smooth(indgen(512,511), 9, /edge_mirror)
>> sm = smooth(indgen(512,510), 9, /edge_mirror)
>>
>> And just to be pedantic, I've tested a few other cases and these did NOT result in a crash:
>> sm = smooth(indgen(512,512), 9, /edge_mirror)
>> sm = smooth(indgen(511,511), 9, /edge_mirror)
>>
>> It appears that smooth (stopped?) being able to smooth rectangular arrays.
>>
>> By running the above from the command line I get the following error (for a couple of seconds on the command line terminal):
>> % Array has a corrupted descriptor: <no name>
>> % Execution halted at: $Main$
>>
>> And my IDL version.
>> IDL> !version
>> {
>> "ARCH": "x86_64",
>> "OS": "Win32",
>> "OS_FAMILY": "Windows",
>> "OS_NAME": "Microsoft Windows",
>> "RELEASE": "8.5.1",
>> "BUILD_DATE": "Nov 14 2015",
>> "MEMORY_BITS": 64,
>> "FILE_OFFSET_BITS": 64
>> }
>>
>> I others are brave enough, could you let me know if it crashes on other versions of IDL? Edge_mirror and _wrap have been added only in 8.1.
>>
>> Cheers,
>> Helder
>
> It also crashes on my Macbook running 8.5.1 and on my Linux systems running (32 bit) 8.2.2 and (64 bit) 8.5.1. So it looks like a problem in the internal SMOOTH() algorithm. --Wayne
>
> "ARCH": "x86_64",
> "OS": "darwin",
> "OS_FAMILY": "unix",
> "OS_NAME": "Mac OS X",
> "RELEASE": "8.5.1",
> "BUILD_DATE": "Nov 14 2015",
> "MEMORY_BITS": 64,
> "FILE_OFFSET_BITS": 64
>
> ARCH STRinG 'x86_64'
> OS STRING 'linux'
> OS_FAMILY STRING 'unix'
> OS_NAME STRING 'linux'
> RELEASE STRING '8.5.1'
> BUILD_DATE STRING 'Nov 14 2015'
> MEMORY_BITS INT 64
> FILE_OFFSET_BITS
> INT 64
>
>
>
> ARCH STRING 'x86'
> OS STRING 'linux'
> OS_FAMILY STRING 'unix'
> OS_NAME STRING 'linux'
> RELEASE STRING '8.2.2'
> BUILD_DATE STRING 'Jan 23 2013'
> MEMORY_BITS INT 32
> FILE_OFFSET_BITS
> INT 64
Hi Wayne,
thanks. I just the info from Harris, that this bug will most likely not make it into 8.6 (January?). It has been logged with ID:IDL-69705.
Thanks for testing.
Cheers,
Helder
|
|
|
Re: smooth() bug in (at least) version 8.5.1 [message #93943 is a reply to message #93939] |
Fri, 02 December 2016 10:38   |
Burch
Messages: 28 Registered: December 2013
|
Junior Member |
|
|
On Friday, December 2, 2016 at 8:03:27 AM UTC-6, Helder wrote:
> Hi,
> I came across this bug and it's a "heavy" one, meaning that IDL crashes and you're thrown back out.
>
> The crashing commands are (try at own risk!):
> sm = smooth(indgen(512,511), 9, /edge_mirror)
> sm = smooth(indgen(512,510), 9, /edge_mirror)
>
> And just to be pedantic, I've tested a few other cases and these did NOT result in a crash:
> sm = smooth(indgen(512,512), 9, /edge_mirror)
> sm = smooth(indgen(511,511), 9, /edge_mirror)
>
> It appears that smooth (stopped?) being able to smooth rectangular arrays.
>
> By running the above from the command line I get the following error (for a couple of seconds on the command line terminal):
> % Array has a corrupted descriptor: <no name>
> % Execution halted at: $Main$
>
> And my IDL version.
> IDL> !version
> {
> "ARCH": "x86_64",
> "OS": "Win32",
> "OS_FAMILY": "Windows",
> "OS_NAME": "Microsoft Windows",
> "RELEASE": "8.5.1",
> "BUILD_DATE": "Nov 14 2015",
> "MEMORY_BITS": 64,
> "FILE_OFFSET_BITS": 64
> }
>
> I others are brave enough, could you let me know if it crashes on other versions of IDL? Edge_mirror and _wrap have been added only in 8.1.
>
> Cheers,
> Helder
It failed for me too. However, I was able to work around it by doing the following:
IDL> arr = indgen(512,511)
IDL> sm1 = smooth(arr, [9,1], /edge_mirror)
IDL> sm2 = smooth(sm1, [1,9], /edge_mirror)
Interestingly, this only failed for me when using integer types. For example, when using floats it worked fine:
IDL> arrFloat = float(arr)
IDL> smFloat = smooth(arrFloat, 9, /edge_mirror)
and to show that I get the same answer as above (at least in this case) by smoothing along each dimension in separate commands:
IDL> smFLoat1 = smooth(arrFloat, [9,1], /edge_mirror)
IDL> smFloat2 = smooth(smFloat1, [1,9], /edge_mirror)
IDL> print, moment(smFloat2 - smFloat)
0.00000 0.00000 NaN NaN
IDL> !version
{
"ARCH": "x86_64",
"OS": "darwin",
"OS_FAMILY": "unix",
"OS_NAME": "Mac OS X",
"RELEASE": "8.5",
"BUILD_DATE": "Jul 7 2015",
"MEMORY_BITS": 64,
"FILE_OFFSET_BITS": 64
}
-Jeff
|
|
|
Re: smooth() bug in (at least) version 8.5.1 [message #93944 is a reply to message #93943] |
Fri, 02 December 2016 10:50   |
Burch
Messages: 28 Registered: December 2013
|
Junior Member |
|
|
On Friday, December 2, 2016 at 12:38:03 PM UTC-6, Jeff B wrote:
> On Friday, December 2, 2016 at 8:03:27 AM UTC-6, Helder wrote:
>> Hi,
>> I came across this bug and it's a "heavy" one, meaning that IDL crashes and you're thrown back out.
>>
>> The crashing commands are (try at own risk!):
>> sm = smooth(indgen(512,511), 9, /edge_mirror)
>> sm = smooth(indgen(512,510), 9, /edge_mirror)
>>
>> And just to be pedantic, I've tested a few other cases and these did NOT result in a crash:
>> sm = smooth(indgen(512,512), 9, /edge_mirror)
>> sm = smooth(indgen(511,511), 9, /edge_mirror)
>>
>> It appears that smooth (stopped?) being able to smooth rectangular arrays.
>>
>> By running the above from the command line I get the following error (for a couple of seconds on the command line terminal):
>> % Array has a corrupted descriptor: <no name>
>> % Execution halted at: $Main$
>>
>> And my IDL version.
>> IDL> !version
>> {
>> "ARCH": "x86_64",
>> "OS": "Win32",
>> "OS_FAMILY": "Windows",
>> "OS_NAME": "Microsoft Windows",
>> "RELEASE": "8.5.1",
>> "BUILD_DATE": "Nov 14 2015",
>> "MEMORY_BITS": 64,
>> "FILE_OFFSET_BITS": 64
>> }
>>
>> I others are brave enough, could you let me know if it crashes on other versions of IDL? Edge_mirror and _wrap have been added only in 8.1.
>>
>> Cheers,
>> Helder
>
> It failed for me too. However, I was able to work around it by doing the following:
>
> IDL> arr = indgen(512,511)
> IDL> sm1 = smooth(arr, [9,1], /edge_mirror)
> IDL> sm2 = smooth(sm1, [1,9], /edge_mirror)
>
> Interestingly, this only failed for me when using integer types. For example, when using floats it worked fine:
>
> IDL> arrFloat = float(arr)
> IDL> smFloat = smooth(arrFloat, 9, /edge_mirror)
>
> and to show that I get the same answer as above (at least in this case) by smoothing along each dimension in separate commands:
>
> IDL> smFLoat1 = smooth(arrFloat, [9,1], /edge_mirror)
> IDL> smFloat2 = smooth(smFloat1, [1,9], /edge_mirror)
> IDL> print, moment(smFloat2 - smFloat)
> 0.00000 0.00000 NaN NaN
>
> IDL> !version
> {
> "ARCH": "x86_64",
> "OS": "darwin",
> "OS_FAMILY": "unix",
> "OS_NAME": "Mac OS X",
> "RELEASE": "8.5",
> "BUILD_DATE": "Jul 7 2015",
> "MEMORY_BITS": 64,
> "FILE_OFFSET_BITS": 64
> }
>
> -Jeff
I'm sure you know this, but I should make clear that the results for floats will NOT be the same as the results for integers. If you input an integer type then smooth will do integer math:
IDL> arr = indgen(5)
IDL> print, smooth(arr, 3, /edge_mirror)
0 1 2 3 3
IDL> print, smooth(float(arr), 3, /edge_mirror)
0.333333 1.00000 2.00000 3.00000 3.66667
-Jeff
|
|
|
Re: smooth() bug in (at least) version 8.5.1 [message #93947 is a reply to message #93943] |
Fri, 02 December 2016 12:36   |
Helder Marchetto
Messages: 520 Registered: November 2011
|
Senior Member |
|
|
On Friday, December 2, 2016 at 7:38:03 PM UTC+1, Jeff B wrote:
> On Friday, December 2, 2016 at 8:03:27 AM UTC-6, Helder wrote:
>> Hi,
>> I came across this bug and it's a "heavy" one, meaning that IDL crashes and you're thrown back out.
>>
>> The crashing commands are (try at own risk!):
>> sm = smooth(indgen(512,511), 9, /edge_mirror)
>> sm = smooth(indgen(512,510), 9, /edge_mirror)
>>
>> And just to be pedantic, I've tested a few other cases and these did NOT result in a crash:
>> sm = smooth(indgen(512,512), 9, /edge_mirror)
>> sm = smooth(indgen(511,511), 9, /edge_mirror)
>>
>> It appears that smooth (stopped?) being able to smooth rectangular arrays.
>>
>> By running the above from the command line I get the following error (for a couple of seconds on the command line terminal):
>> % Array has a corrupted descriptor: <no name>
>> % Execution halted at: $Main$
>>
>> And my IDL version.
>> IDL> !version
>> {
>> "ARCH": "x86_64",
>> "OS": "Win32",
>> "OS_FAMILY": "Windows",
>> "OS_NAME": "Microsoft Windows",
>> "RELEASE": "8.5.1",
>> "BUILD_DATE": "Nov 14 2015",
>> "MEMORY_BITS": 64,
>> "FILE_OFFSET_BITS": 64
>> }
>>
>> I others are brave enough, could you let me know if it crashes on other versions of IDL? Edge_mirror and _wrap have been added only in 8.1.
>>
>> Cheers,
>> Helder
>
> It failed for me too. However, I was able to work around it by doing the following:
>
> IDL> arr = indgen(512,511)
> IDL> sm1 = smooth(arr, [9,1], /edge_mirror)
> IDL> sm2 = smooth(sm1, [1,9], /edge_mirror)
>
> Interestingly, this only failed for me when using integer types. For example, when using floats it worked fine:
>
> IDL> arrFloat = float(arr)
> IDL> smFloat = smooth(arrFloat, 9, /edge_mirror)
>
> and to show that I get the same answer as above (at least in this case) by smoothing along each dimension in separate commands:
>
> IDL> smFLoat1 = smooth(arrFloat, [9,1], /edge_mirror)
> IDL> smFloat2 = smooth(smFloat1, [1,9], /edge_mirror)
> IDL> print, moment(smFloat2 - smFloat)
> 0.00000 0.00000 NaN NaN
>
> IDL> !version
> {
> "ARCH": "x86_64",
> "OS": "darwin",
> "OS_FAMILY": "unix",
> "OS_NAME": "Mac OS X",
> "RELEASE": "8.5",
> "BUILD_DATE": "Jul 7 2015",
> "MEMORY_BITS": 64,
> "FILE_OFFSET_BITS": 64
> }
>
> -Jeff
Thanks! This saves the day... or something like that.
:-)
Cheers,
Helder
|
|
|
Re: smooth() bug in (at least) version 8.5.1 [message #93964 is a reply to message #93947] |
Wed, 07 December 2016 12:25  |
chris_torrence@NOSPAM
Messages: 528 Registered: March 2007
|
Senior Member |
|
|
On Friday, December 2, 2016 at 1:36:47 PM UTC-7, Helder wrote:
> On Friday, December 2, 2016 at 7:38:03 PM UTC+1, Jeff B wrote:
>> On Friday, December 2, 2016 at 8:03:27 AM UTC-6, Helder wrote:
>>> Hi,
>>> I came across this bug and it's a "heavy" one, meaning that IDL crashes and you're thrown back out.
>>>
>>> The crashing commands are (try at own risk!):
>>> sm = smooth(indgen(512,511), 9, /edge_mirror)
>>> sm = smooth(indgen(512,510), 9, /edge_mirror)
>>>
>>> And just to be pedantic, I've tested a few other cases and these did NOT result in a crash:
>>> sm = smooth(indgen(512,512), 9, /edge_mirror)
>>> sm = smooth(indgen(511,511), 9, /edge_mirror)
>>>
>>> It appears that smooth (stopped?) being able to smooth rectangular arrays.
>>>
>>> By running the above from the command line I get the following error (for a couple of seconds on the command line terminal):
>>> % Array has a corrupted descriptor: <no name>
>>> % Execution halted at: $Main$
>>>
>>> And my IDL version.
>>> IDL> !version
>>> {
>>> "ARCH": "x86_64",
>>> "OS": "Win32",
>>> "OS_FAMILY": "Windows",
>>> "OS_NAME": "Microsoft Windows",
>>> "RELEASE": "8.5.1",
>>> "BUILD_DATE": "Nov 14 2015",
>>> "MEMORY_BITS": 64,
>>> "FILE_OFFSET_BITS": 64
>>> }
>>>
>>> I others are brave enough, could you let me know if it crashes on other versions of IDL? Edge_mirror and _wrap have been added only in 8.1.
>>>
>>> Cheers,
>>> Helder
>>
>> It failed for me too. However, I was able to work around it by doing the following:
>>
>> IDL> arr = indgen(512,511)
>> IDL> sm1 = smooth(arr, [9,1], /edge_mirror)
>> IDL> sm2 = smooth(sm1, [1,9], /edge_mirror)
>>
>> Interestingly, this only failed for me when using integer types. For example, when using floats it worked fine:
>>
>> IDL> arrFloat = float(arr)
>> IDL> smFloat = smooth(arrFloat, 9, /edge_mirror)
>>
>> and to show that I get the same answer as above (at least in this case) by smoothing along each dimension in separate commands:
>>
>> IDL> smFLoat1 = smooth(arrFloat, [9,1], /edge_mirror)
>> IDL> smFloat2 = smooth(smFloat1, [1,9], /edge_mirror)
>> IDL> print, moment(smFloat2 - smFloat)
>> 0.00000 0.00000 NaN NaN
>>
>> IDL> !version
>> {
>> "ARCH": "x86_64",
>> "OS": "darwin",
>> "OS_FAMILY": "unix",
>> "OS_NAME": "Mac OS X",
>> "RELEASE": "8.5",
>> "BUILD_DATE": "Jul 7 2015",
>> "MEMORY_BITS": 64,
>> "FILE_OFFSET_BITS": 64
>> }
>>
>> -Jeff
>
> Thanks! This saves the day... or something like that.
> :-)
> Cheers,
> Helder
Hi all,
Yes, this bug has been in the SMOOTH function since IDL 8.1 when the EDGE_MIRROR and EDGE_WRAP keywords were added. It works fine for square arrays, but walks off the end of the array for non-square arrays.
It's now been fixed, and will be in IDL Next (not 8.6).
Cheers,
Chris
Harris Geospatial Solutions
|
|
|