On Wed, 14 Jan 2009 18:31:14 -0600, n@@dles wrote:
>Apologies I messed up here :( I have your csvs for 21 and 29 and baldricks
>for 26.
>
My csv for issue 26 should be in the 'bible', but I'll attach another copy
here for you.
>> If you would, please look the two ga- files over, compare them with
>> yours, and let me know what you think.
>>
>
>I looked at both copies and as you say the headers match for the images.
>
>The CD versions of the images although less corrupt have filesizes
>considerably less than the average for the set. i.e ga-196-50 is 151k and
>no other image in that set is below 171k with the majority being 200k+.
>The version i have of that image is 249k which fits more with the images
>sizes.
>It leads me to believe that someone has attempted to fix the corruption
>which has reduced the image size.
>
>This is really subjective tho and since both versions of the images are
>essentially bad because of the corruption i would say it is up to the
>individual.
>
>I did notice that the CD version of the thumbnail doesnt have a matching
>header tho.
>
>Personally I am inclined to go with the versions i already have because of
>the above but also in these sort of situations I know that it was my own
>rip that these came from whereas the CD versions are from an unknown source
>:)
>
>What do you think??
>
I guess I wasn't aware they were of two differing sizes until you pointed it
out. I would normally expect such corruption to come from a truncated file.
In that case, being too small is not of concern since the non-truncated
version of the file would be larger.
When I looked at both with a hex editor I was struck by their similarity. In
fact, they were byte-for-byte identical up to the point of corruption of your
file. The real difference was that your file was not truncated, it had an
end of image marker, whereas CD's file was truncated. However, CD's file had
good image bytes up to the point of truncation, yours did not stay good that
deep into the image.
That's when I got the bright idea to copy the extended good portion of CD's
file and allow it to overwrite that same corrupt portion of yours. My hope
was that maybe the corrupt area would get overwritten with good bytes at
which point maybe there wouldn't still be any corruption. So in effect, the
whole file could be repaired.
Alas, not to be! The image simply became good a bit further, enough to match
the area of good in CD's version, but the corruption immediately set in
again. Thinking there may have been a loss of sync in transmission, I
searched both copies of the files for short sequences of bytes in the other.
There was no point where the two files had even a tiny amount of congruity
past the point of corruption in your image. There was no way to put the
files in sync and create a full image.
My next guess was that the original truncated file had been on a crashed hard
drive, and that during recovery it had gotten cross-linked with another file,
thus ending up with an end of image marker. So I searched all the other
files in the folder for a short sequence of bytes found in the corrupt area
of your copy of the file. Nothing!
My last attempt at sanity was to measure the height of the image frame on my
monitor, and the height of the good part of the image as presented by CD's
version, then worked out the ratio of good to bad and calculated a presumed
full file size for a complete file. It matched yours within 4k but it seems
evident to me your copy contains bytes from some other jpeg.
I'm stymied.
I checked the other file too. They both remain congruent up to the point of
corruption, which is very deep in the file. CD's file finishes off fine, no
corruption in the horizontal, and has an end of image marker.
Your copy goes corrupt in the horizontal near the end. It too has an end of
image marker, but is a much larger file than CD's - remember the two are
congruent to the point of corruption, so it stands to reason it too has the
ending bytes from some other jpeg.
Now I'm really stymied!
>Regards
>n@@dles
>
>
--
mr.bill
|
|