Red Hat Bugzilla – Bug 232785
jhead inserts to the images metadata random non-printable characters
Last modified: 2007-11-30 17:11:59 EST
Description of problem: [matej@chelcicky 18]$ jhead -dc p20041118_132838.jpg [matej@chelcicky 18]$ jhead p20041118_132838.jpg File name : p20041118_132838.jpg File size : 1754224 bytes File date : 2004:11:18 13:28:38 Camera make : SONY Camera model : DSC-P73 Date/Time : 2004:11:18 13:28:38 Resolution : 1728 x 2304 Flash used : Yes (auto, return light detected) Focal length : 6.0mm Exposure time: 0.025 s (1/40) Aperture : f/2.8 ISO equiv. : 100 Whitebalance : Auto Metering Mode: matrix Exposure : program (auto) Comment : (the last character in comment, if it is not preserved in bugzilla, is chr(2)) The character is not only chr(2), but also chr(1), chr(7), "d", and other characters. The operation which corrupts metadata is not only -dc but different operations with the image do it as well. Version-Release number of selected component (if applicable): jhead-2.6-3.fc6 How reproducible: from my sample of 1700+ images, I get corrupted some couple of hundred of them. Tried many times to remove garbage characters (jhead -cl seems to not introduce anything else, so attached script works for me) and always got couple of hundred corrupted image files. Steps to Reproduce: see above or do other operations on the images with jhead. Actual results: by using jhead on my images random characters are introduced to the images metadata. Expected results: jhead should not take its own initiative about the content of my metadata
Created attachment 150318 [details] image with corrupted metadata
Created attachment 150320 [details] script mentioned in the comment
Still reproducible with jhead-2.7-1.fc6
I thought that the new release will probably not help but I tried it... just in case. I will try to get in contact with upstream to see if they have any idea what might be wrong here.
Following question from the author: Well, there's nonprintable characters in the exif comment. I guess I should filter those out if present. Where did this image come from? Just curious how that comment got inserted. Also curious what the best strategy should be in terms of dealing with that kind of corruption. And: I'll work on something to clean this up. Just more of a question where the image came from so I get a better sense of what else I might expect. Its specific to the image and whatever tool processed it at any rate, so its not a major issue. You can either directly contact the author or just let me know here and I forward it.
Created attachment 150521 [details] another example of corrupted image One more attempt to get rid of the non-printable characters. [matej@chelcicky 18]$ jhead p20041118_132838.jpg File name : p20041118_132838.jpg File size : 1754224 bytes File date : 2004:11:18 13:28:38 Camera make : SONY Camera model : DSC-P73 Date/Time : 2004:11:18 13:28:38 Resolution : 1728 x 2304 Flash used : Yes (auto, return light detected) Focal length : 6.0mm Exposure time: 0.025 s (1/40) Aperture : f/2.8 ISO equiv. : 100 Whitebalance : Auto Metering Mode: matrix Exposure : program (auto) Comment : [matej@chelcicky 18]$ jhead -cl "" p20041118_132838.jpg Comment not modified [matej@chelcicky 18]$ jhead p20041118_132838.jpg File name : p20041118_132838.jpg File size : 1754224 bytes File date : 2004:11:18 13:28:38 Camera make : SONY Camera model : DSC-P73 Date/Time : 2004:11:18 13:28:38 Resolution : 1728 x 2304 Flash used : Yes (auto, return light detected) Focal length : 6.0mm Exposure time: 0.025 s (1/40) Aperture : f/2.8 ISO equiv. : 100 Whitebalance : Auto Metering Mode: matrix Exposure : program (auto) Comment : [matej@chelcicky 18]$ exiftran -c "" p20041118_132838.jpg you have to either specify a output file (-o <file>) or enable inplace editing (-i). Try -h for more info. [matej@chelcicky 18]$ exiftran -c "" -i p20041118_132838.jpg processing p20041118_132838.jpg [matej@chelcicky 18]$ jhead p20041118_132838.jpg File name : p20041118_132838.jpg File size : 1754247 bytes File date : 2007:03:20 19:44:18 Camera make : SONY Camera model : DSC-P73 Date/Time : 2004:11:18 13:28:38 Resolution : 1728 x 2304 Flash used : Yes (auto, return light detected) Focal length : 6.0mm Exposure time: 0.025 s (1/40) Aperture : f/2.8 ISO equiv. : 100 Whitebalance : Auto Metering Mode: matrix Exposure : program (auto) Comment : ? [matej@chelcicky 18]$ jhead -cl "" p20041118_132838.jpg Comment not modified [matej@chelcicky 18]$ Otherwise, to upstream maintainer's questions (are we talking about Matthias Wandel? Shouldn't we added him to Cc: of this bug?): 1) I don't think it is matter of one image -- I have hundreds of images with comments like this (see above). This one is originally from my camera (Sony Cybershot DSC-P73) and went through a lot of programs -- for the longest time it was imported from the camera and then managed with digikam, then for the brief time I tried f-spot, but now I switched to jbrout, which uses jhead and exiftran internally for all its image processing (and I am applying for review in Fedora Extras -- Adrian, wouldn't you be so kind and make review of bug 230316? :-)) I am almost certain, that these characters were not present in the file until it was touched by jbrout (and thus by jhead), but I couldn't prove it. I am 100% sure, that it was OK in digikam, I tried f-spot so briefly (Novell just then announced deal with Microsoft so I got rid of all Mono-based apps then) that I am not sure. I could try, but I don't have any idea how to triage it. 2) How to deal with these characters? I don't know. Well, first of all, it would be useful to find out how (and whether) either jhead or exiftran really corrupt metadata and fix it ;-). However, concerning the already corrupted files, I am not sure -- I could certainly make some script to delete all one character comments (if jhead actually worked). See this another attached image -- comment here is only letter "d", so even filtering on non-printable characters (although it would be quite useful) wouldn't help. BTW, this image was created with another camera and nothing else than gthum-import, exiftrans (possibly) and jhead haven't touched it.
Please contact Matthias Wandel because I cannot help you here. He does not want to create yet another account to have bugzilla access. I have received another mail from him: I'm not sure if its wise to fix. What if the strings are in UTF8? In that case, I shouldn't discard "nonprintable" characters, should I? Maybe its best to close this as a non-bug Grep also outputs nonprintable characters if the source data has nonprintable characters. I am closing the bug with UPSTREAM because that is the only way it can be resolved. Sorry cannot help you here.