Interesting. Could you try to open the file with the exif functions?

But however the file is borked.

/usr/local/ImageMagick/bin/identify -format "%[EXIF:*]" wrong_jpeg.jpg
Make=Konica
Model=Revio C2
Orientation=1
XResolution=288/3
YResolution=288/3
ResolutionUnit=2
DateTime=2003:10:15 13:48:10
YCbCrPositioning=2
ExifOffset=174
ExposureTime=5924356/268435456
FNumber=28/10
ExposureProgram=3
ISOSpeedRatings=64
ExifVersion=0220
DateTimeOriginal=2003:10:15 13:48:10
DateTimeDigitized=2003:10:15 13:48:10
ComponentsConfiguration=...
CompressedBitsPerPixel=1989456/1228800
ShutterSpeedValue=44/8
ApertureValue=28/10
ExposureBiasValue=0/10
MaxApertureValue=28/10
SubjectDistance=11/10
MeteringMode=5
LightSource=0
Flash=0
FocalLength=45/10
MakerNote=0060162454
FlashPixVersion=0100
ColorSpace=1
ExifImageWidth=1280
ExifImageLength=960
unknown=
InteroperabilityOffset=724
unknown=R98
unknown=0100
ExposureIndex=1/1
SensingMethod=2
FileSource=.
SceneType=.
unknown=1280/1280
unknown=37

jpeg lib also working properly with this image.
Please see this code (from jpeg-6b):
next_marker (j_decompress_ptr cinfo)
{
  int c;
  INPUT_VARS(cinfo);

  for (;;) {
    INPUT_BYTE(cinfo, c, return FALSE);
    /* Skip any non-FF bytes. 
     * This may look a bit inefficient, but it will not occur in a valid file. 
     * We sync after each discarded byte so that a suspending data source 
     * can discard the byte from its buffer. 
     */
    while (c != 0xFF) {
      cinfo->marker->discarded_bytes++;
      INPUT_SYNC(cinfo);
      INPUT_BYTE(cinfo, c, return FALSE);
    }
    /* This loop swallows any duplicate FF bytes.  Extra FFs are legal as 
     * pad bytes, so don't count them in discarded_bytes.  We assume there 
     * will not be so many consecutive FF bytes as to overflow a suspending 
     * data source's input buffer. 
     */
    do {
      INPUT_BYTE(cinfo, c, return FALSE);
    } while (c == 0xFF);
    if (c != 0)
      break;            /* found a valid marker, exit loop */
    /* Reach here if we found a stuffed-zero data sequence (FF/00). 
     * Discard it and loop back to try again. 
     */
    cinfo->marker->discarded_bytes += 2;
    INPUT_SYNC(cinfo);
  }
  
  if (cinfo->marker->discarded_bytes != 0) {
    WARNMS2(cinfo, JWRN_EXTRANEOUS_DATA, cinfo->marker->discarded_bytes, c);
    cinfo->marker->discarded_bytes = 0;
  }
  
  cinfo->unread_marker = c;

  INPUT_SYNC(cinfo);
  return TRUE;
}

Sorry, but your problem does not imply a bug in PHP itself.  For a
list of more appropriate places to ask for help using PHP, please
visit http://www.php.net/support.php as this bug system is not the
appropriate forum for asking support questions. 

Thank you for your interest in PHP.

Your image file is corrupt and i don\'t think i will implement a special handling for all wrong software out there. That would lead to unrobust php code.

ok, why you implement this ?

/* get marker byte, swallowing possible padding */
if ( last_marker==M_COM && comment_correction) {
/* some software does not count the length bytes of COM section */
/* one company doing so is very much envolved in JPEG... so we accept too */
/* by the way: some of those companies changed their code now... */
  comment_correction = 2;
} else {
  last_marker = 0;
  comment_correction = 0;
}

Well it's a valid point you can take. Probably i'll even adapt the code for getImageSize() but not now.

reopen

The code you mentioned is to support the 'standard' error in writing wrong jpeg files. I do not plan to add any additional workarounds for any other software whose manufacturer cannot read a standard. If you feel a need for this go ahead and show me a working patch. If that does not hurt robustness too much, i will consider applying it.

Also see: Bug #13213 Unknown image format