Bug #102118 for Image-ExifTool: t/Nikon.t test fails with -Duselongdouble (only FreeBSD)

Sat Feb 14 02:52:38 2015 SREZIC [...] cpan.org - Ticket created

Subject:

t/Nikon.t test fails with -Duselongdouble (only FreeBSD)

On a FreeBSD 10.1 system with perl 5.20.1 with uselongdouble enabled "make test" fails: Test 4 differs beginning at line 93: Test gave: "[MakerNotes, Nikon, Camera] 132 - Lens: 18-70mm f/3.50000000000000022-4.50000000000000028" Should be: "[MakerNotes, Nikon, Camera] 132 - Lens: 18-70mm f/3.5-4.5" Test 5 differs beginning at line 103: Test gave: "[MakerNotes, Nikon, Camera] 132 - Lens: 50mm f/1.80000000000000011" Should be: "[MakerNotes, Nikon, Camera] 132 - Lens: 50mm f/1.8" Test 7 differs beginning at line 111: Test gave: "[MakerNotes, Nikon, Camera] 132 - Lens: 18-70mm f/3.50000000000000022-4.50000000000000028" Should be: "[MakerNotes, Nikon, Camera] 132 - Lens: 18-70mm f/3.5-4.5" t/Nikon.t .......... Failed 3/9 subtests This seems to be the only platform where the test is failing... linux with -Duselongdouble seems to be fine.

Sat Feb 14 08:03:47 2015 phil [...] owl.phy.queensu.ca - Correspondence added

CC:	Phil Harvey <phil [...] physics.queensu.ca>
Subject:	Re: [rt.cpan.org #102118] t/Nikon.t test fails with -Duselongdouble (only FreeBSD)
Date:	Sat, 14 Feb 2015 08:02:21 -0500
To:	SREZIC [...] cpan.org, bug-Image-ExifTool [...] rt.cpan.org
From:	Phil Harvey <phil [...] owl.phy.queensu.ca>

Thanks for this report. Would it be possible for you to look at the resulting t/Nikon_4.failed file to see if Lens is the only effected tag? My tests will only report the first occurrence in each file. The output of this command will give me all the information I need: diff t/Nikon_4.out t/Nikon_4.failed Thanks. - Phil On Feb 14, 2015, at 2:52 AM, Slaven_Rezic via RT wrote: Show quoted text

> Test 4 differs beginning at line 93: > Test gave: "[MakerNotes, Nikon, Camera] 132 - Lens: 18-70mm > f/3.50000000000000022-4.50000000000000028" > Should be: "[MakerNotes, Nikon, Camera] 132 - Lens: 18-70mm > f/3.5-4.5"

Sat Feb 14 08:03:48 2015 The RT System itself - Status changed from 'new' to 'open'

Sat Feb 14 12:14:48 2015 SREZIC [...] cpan.org - Correspondence added

On 2015-02-14 08:03:47, phil@owl.phy.queensu.ca wrote: Show quoted text

> Thanks for this report. > > Would it be possible for you to look at the resulting t/Nikon_4.failed > file to see if Lens is the only effected tag? My tests will only report > the first occurrence in each file. > > The output of this command will give me all the information I need: > > diff t/Nikon_4.out t/Nikon_4.failed > > Thanks. > > - Phil > > On Feb 14, 2015, at 2:52 AM, Slaven_Rezic via RT wrote:

> > Test 4 differs beginning at line 93: > > Test gave: "[MakerNotes, Nikon, Camera] 132 - Lens: 18-70mm > > f/3.50000000000000022-4.50000000000000028" > > Should be: "[MakerNotes, Nikon, Camera] 132 - Lens: 18-70mm > > f/3.5-4.5"

>

It seems that there are more differences --- diff attached.

Subject:

nikon.diff

1c1 < [ExifTool, ExifTool, ExifTool] ExifToolVersion - ExifTool Version Number: 9.32 --- > [ExifTool, ExifTool, ExifTool] ExifToolVersion - ExifTool Version Number: 9.76 4,8c4,8 < [File, System, Image] FileSize - File Size: 6.8 kB < [File, System, Time] FileModifyDate - File Modification Date/Time: 2013:06:12 12:34:49-04:00 < [File, System, Time] FileAccessDate - File Access Date/Time: 2013:06:12 12:34:49-04:00 < [File, System, Time] FileInodeChangeDate - File Inode Change Date/Time: 2013:06:12 12:34:49-04:00 < [File, System, Image] FilePermissions - File Permissions: rw-r--r-- --- > [File, System, Image] FileSize - File Size: 6.9 kB > [File, System, Time] FileModifyDate - File Modification Date/Time: 2015:02:14 18:13:22+01:00 > [File, System, Time] FileAccessDate - File Access Date/Time: 2015:02:14 18:13:22+01:00 > [File, System, Time] FileInodeChangeDate - File Inode Change Date/Time: 2015:02:14 18:13:22+01:00 > [File, System, Image] FilePermissions - File Permissions: rw-rw-r-- 26,27c26,27 < [EXIF, IFD0, Camera] 318 - White Point: 0.313 0.329 < [EXIF, IFD0, Image] 319 - Primary Chromaticities: 0.64 0.33 0.21 0.71 0.15 0.06 --- > [EXIF, IFD0, Camera] 318 - White Point: 0.313000000000000011 0.329000000000000012 > [EXIF, IFD0, Image] 319 - Primary Chromaticities: 0.640000000000000023 0.330000000000000012 0.210000000000000008 0.710000000000000026 0.150000000000000005 0.0600000000000000061 66c66 < [EXIF, ExifIFD, Image] 42240 - Gamma: 2.2 --- > [EXIF, ExifIFD, Image] 42240 - Gamma: 2.20000000000000014 93c93 < [MakerNotes, Nikon, Camera] 132 - Lens: 18-70mm f/3.5-4.5 --- > [MakerNotes, Nikon, Camera] 132 - Lens: 18-70mm f/3.50000000000000022-4.50000000000000028 120c120 < [MakerNotes, Nikon, Camera] 154 - Sensor Pixel Size: 7.8 x 7.8 um --- > [MakerNotes, Nikon, Camera] 154 - Sensor Pixel Size: 7.80000000000000049 x 7.80000000000000049 um 139c139 < [XMP, XMP-x, Document] xmptk - XMP Toolkit: Image::ExifTool 9.32 --- > [XMP, XMP-x, Document] xmptk - XMP Toolkit: Image::ExifTool 9.76 141c141 < [XMP, XMP-aux, Camera] Lens - Lens: 18-70mm f/3.5-4.5 --- > [XMP, XMP-aux, Camera] Lens - Lens: 18-70mm f/3.50000000000000022-4.50000000000000028 160c160 < [Composite, Composite, Camera] LensSpec - Lens: 18-70mm f/3.5-4.5 G --- > [Composite, Composite, Camera] LensSpec - Lens: 18-70mm f/3.50000000000000022-4.50000000000000028 G

Sat Feb 14 18:35:01 2015 phil [...] owl.phy.queensu.ca - Correspondence added

CC:	Phil Harvey <phil [...] owl.phy.queensu.ca>
Subject:	Re: [rt.cpan.org #102118] t/Nikon.t test fails with -Duselongdouble (only FreeBSD)
Date:	Sat, 14 Feb 2015 18:33:14 -0500
To:	SREZIC [...] cpan.org, bug-Image-ExifTool [...] rt.cpan.org
From:	Phil Harvey <phil [...] owl.phy.queensu.ca>

Thanks, This is unfortunate, and what I had feared. To fix this problem at my end I would need to patch the output formatting for all tags which are derived from rational values. Below, for example, the last tag (SensorPixelSize) is calculated simply as 78/10. I personally would consider it a Perl bug if Perl on this system prints 7.80000000000000049 instead of 7.8 with "print 78/10". Is this the case? If so, then I really think Perl needs to be fixed. - Phil On Feb 14, 2015, at 12:14 PM, Slaven_Rezic via RT wrote: Show quoted text

> It seems that there are more differences --- diff attached. >

> > [EXIF, IFD0, Camera] 318 - White Point: 0.313000000000000011

> 0.329000000000000012

> > [EXIF, IFD0, Image] 319 - Primary Chromaticities:

> 0.640000000000000023 0.330000000000000012 0.210000000000000008 > 0.710000000000000026 0.150000000000000005 0.0600000000000000061

> > [EXIF, ExifIFD, Image] 42240 - Gamma: 2.20000000000000014 > > [MakerNotes, Nikon, Camera] 132 - Lens: 18-70mm

> f/3.50000000000000022-4.50000000000000028

> > [MakerNotes, Nikon, Camera] 154 - Sensor Pixel Size:

> 7.80000000000000049 x 7.80000000000000049 um

Sun Feb 15 04:55:19 2015 SREZIC [...] cpan.org - Correspondence added

On 2015-02-14 18:35:01, phil@owl.phy.queensu.ca wrote: Show quoted text

> Thanks, > > This is unfortunate, and what I had feared. To fix this problem at my > end I would > need to patch the output formatting for all tags which are derived from > rational values. > Below, for example, the last tag (SensorPixelSize) is calculated simply > as 78/10. > > I personally would consider it a Perl bug if Perl on this system prints > 7.80000000000000049 instead of 7.8 with "print 78/10". Is this the > case?

No, this looks fine. On this freebsd 10.1 system: $ perl5.20.1D -e 'print 78/10, "\n"' 7.8 It starts to look strange when using printf with 19 or more digits after the decimal point: $ perl5.20.1D -e 'printf "%.19f\n", 78/10' 7.8000000000000000002 But this is also the case on a linux platform with a -Duselongdouble perl. And a "normal" perl looks even worse with this precision: $ perl5.20.1 -e 'printf "%.19f\n", 78/10' 7.7999999999999998224 Show quoted text

> If so, then I really think Perl needs to be fixed. > > - Phil > > On Feb 14, 2015, at 12:14 PM, Slaven_Rezic via RT wrote:

> > It seems that there are more differences --- diff attached. > >

> > > [EXIF, IFD0, Camera] 318 - White Point: 0.313000000000000011

> > 0.329000000000000012

> > > [EXIF, IFD0, Image] 319 - Primary Chromaticities:

> > 0.640000000000000023 0.330000000000000012 0.210000000000000008 > > 0.710000000000000026 0.150000000000000005 0.0600000000000000061

> > > [EXIF, ExifIFD, Image] 42240 - Gamma: 2.20000000000000014 > > > [MakerNotes, Nikon, Camera] 132 - Lens: 18-70mm

> > f/3.50000000000000022-4.50000000000000028

> > > [MakerNotes, Nikon, Camera] 154 - Sensor Pixel Size:

> > 7.80000000000000049 x 7.80000000000000049 um

>

Sun Feb 15 09:18:16 2015 phil [...] owl.phy.queensu.ca - Correspondence added

Subject:	Re: [rt.cpan.org #102118] t/Nikon.t test fails with -Duselongdouble (only FreeBSD)
Date:	Sun, 15 Feb 2015 09:16:35 -0500
To:	bug-Image-ExifTool [...] rt.cpan.org
From:	Phil Harvey <phil [...] owl.phy.queensu.ca>

Thanks for all your help. I have looked into this in more detail, and found that the problem is likely at my end. I will apply a patch, and hopefully this will be fixed in ExifTool 9.86 and later. - Phil

Mon Mar 16 10:44:57 2015 EXIFTOOL [...] cpan.org - Status changed from 'open' to 'resolved'

Mon Mar 16 10:44:58 2015 EXIFTOOL [...] cpan.org - Fixed in 9.90 added

Mon Mar 16 16:37:15 2015 SREZIC [...] cpan.org - Correspondence added

On 2015-02-15 09:18:16, phil@owl.phy.queensu.ca wrote: Show quoted text

> Thanks for all your help. > > I have looked into this in more detail, and found that the problem > is likely at my end. I will apply a patch, and hopefully this will be > fixed in ExifTool 9.86 and later.

Looks good in 9.90 --- test suite passes with my -Duselongdouble perl.