Your message dated Mon, 26 Oct 2026 18:54:01 +0100
with message-id <[email protected]>
and subject line
has caused the Debian Bug report #1152141,
regarding Rotation by 180° mutilates the columns on the border. Errors in the
manpage.
to be marked as done.
This means that you claim that the problem has been dealt with.
If this is not the case it is now your responsibility to reopen the
Bug report if necessary, and/or fix the problem forthwith.
(NB: If you are a system administrator and have no idea what this
message is talking about, this may indicate a serious mail system
misconfiguration somewhere. Please contact [email protected]
immediately.)
--
1152141: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1152141
Debian Bug Tracking System
Contact [email protected] with problems
--- Begin Message ---
Package: exiftran
Version: 2.14-1
Control: affects -1 loupe jpegtran
Grab a particular JPEG file, crop away the white margins a bit, and rotate the
cropped image by 180°:
wget -O 1.jpg
"https://static.toom.de/produkte/bilder/9191028/9191028_3.png?width=1050&height=1050&format=jpg"
jpegtran -crop 970x794+32+128 -perfect -outfile 2.jpg -strict 1.jpg
exiftran -1 -o 3.jpg 2.jpg
loupe 1.jpg 2.jpg 3.jpg &
Observe the difference between 2.jpg and 3.jpg. The image has been somehow
rotated, but not quite. While 2.jpg does show a lamp in full and small white
margins on all sides, its rotated version in 3.jpg doesn't have a white margin
to the left of the label of the lamp but does have some artifacts on the right
side, to the right of the white margin. (It looks like 10 rightmost columns
might have been flipped upside down instead of being rotated around the center
by 180°, though I didn't check it pixelwise.) This is wrong. No relevant
command (jpegtran, exiftran, and loupe) issued an error or even a warning.
I don't know whose fault it is, so I am issuing a bug report against exiftran
(since the error gets visible with loupe afterward), and the maintainer may
reassign if necessary. Not sure whether loupe is doing wrong itself, but if it
is, so are firefox, display, and gimp, since you can see that something is
wrong when viewing 3.jpg with any of these three commands. Simply changing the
orientation tag in the metadata should not mutilate the image itself, so either
the error silently happened during cropping, or exiftran does more than just
changing the metadata, or the viewers are doing a lousy job afterward.
A possibly related, closed bug: https://bugs.debian.org/289667 .
Further, in the manpage of exiftran we see, “By default exiftran don't preserve
image size of the images that do not meet a multiple of 8 pixels. He prefers to
cut a strip of a few pixels rather than offering a damaged image.” First,
there's a trivial grammatical error here: “exiftran don't” should read
“exiftran doesn't”, and “He” should read “It” or better “Exiftran”. Second,
“multiple” should read “integer multiple”, otherwise the phrase is void of
meaning (e.g., 3 = 0.375 ⋅ 8). Third, both 2.jpg and 3.jpg have the same size,
970×794, and neither of these two dimensions nor their product is an integer
multiple of 8. According to the manpage, the image size of 2.jpg is not
preserved, so the image size of 3.jpg should be different from that of 2.jpg.
Maybe the quotation in the manpage is too strong and should read, “By default,
exiftran retains the image size if the dimensions are integer multiples of 8.
If the dimensions are not integer multiples of 8, exiftran shrinks the image or
leaves it of the same size. Exiftran prefers to cut a strip of a few pixels
rather than offering a damaged image.” Also check whether you want each
dimension or their product to be an integer multiple of 8 and whether you wish
to refer to the dimensions of an MCU block instead of 8.
--- End Message ---
--- Begin Message ---
Replaced by http://bugs.debian.org/1152194 due to a failing retitle command.
--- End Message ---