Over-optimistic detection of file type

[Tom Womack]

Binary package hint: nautilus

I have a text file which begins 'P13438 = 2 * 3 * 7 * 1047137401 * 8868717860281682366053088136263931781236052676953 * 37740643638981149828904962487908210028313437770048087043525652181' - it's a list of factorizations of partition numbers.

For some reason nautilus thinks this is a PBM image, and won't by default open it with gedit.

I think that files which claim to be text files should be opened with the text editor whether or not nautilus thinks they are text files.

[Christophe Fergeau (teuf-gnome)]

So, this was indeed a mismatch, I committed a fix to xdgmime CVS, but now it
will identify files beginning with P1 as portable bitmap files, I guess you
won't consider this bug as fixed in a satisfactory way :)

[Bastien Nocera (hadess-deactivatedaccount)]

$ echo "P1" > foo
$ file foo
foo: Netpbm PBM image text

Should we leave this bug opened?

[Tom Womack (tom-womack)]

Binary package hint: nautilus

I have a text file which begins 'P13438 = 2 * 3 * 7 * 1047137401 * 8868717860281682366053088136263931781236052676953 * 37740643638981149828904962487908210028313437770048087043525652181' - it's a list of factorizations of partition numbers.

For some reason nautilus thinks this is a PBM image, and won't by default open it with gedit.

I think that files which claim to be text files should be opened with the text editor whether or not nautilus thinks they are text files.

[Sebastien Bacher (seb128)]

Thank you for your bug. That's due to the shared-mime-info pbm definition

[Benjamin-close (benjamin-close)]

Bugzilla Upgrade Mass Bug Change

NEEDSINFO state was removed in Bugzilla 3.x, reopening any bugs previously listed as NEEDSINFO.

  - benjsc
    fd.o Wrangler

[Daniel Leidert (dleidert-deactivatedaccount)]

(In reply to comment #0)
> I have a text file that begins with "TP1" (you can create one with just these
> three characters). When double-clicking on it in nautilus, it tells me that the
> filename suggests it's a text file but the file content suggests it's a portable
> bitmap file.

Hm. ATM in fast detection its detected as application/octet-stream by gnomevfs-info. If it would be detected as text/plain, then gnomevfs shouldn't detect it as image/x-portable-bitmap.

I'm not sure if the problem can be solved in the shared-mime-info database. IMO there are two alternatives: 1) remove the magic from image/x-portable-bitmap (no conflicting global pattern) or b) forward this to the upstream of gnomevfs and kdemime and check, if some better heuristics between text/plain and application/octet-stream can solve this problem.

[Daniel Leidert (dleidert-deactivatedaccount)]

Hm. After reading some more about the formats, it should also be possible to extend the graphics a bit to be

P1\x0a# (offset: 0)

so the magic includes the (possible) comment in line 2, although this comment is not required. But it is less generic than just searching for P1..7.

[Sense Egbert Hofstede (sense)]

This bug has also been reported in launchpad for PBM images by Tom Womack: https://bugs.edge.launchpad.net/ubuntu/+source/shared-mime-info/+bug/99736
He said:
"I have a text file which begins 'P13438 = 2 * 3 * 7 * 1047137401 * 8868717860281682366053088136263931781236052676953 * 37740643638981149828904962487908210028313437770048087043525652181' - it's a list of factorizations of partition numbers.

For some reason nautilus thinks this is a PBM image, and won't by default open it with gedit.

I think that files which claim to be text files should be opened with the text editor whether or not nautilus thinks they are text files."

[Sense Egbert Hofstede (sense)]

Comment added to the freedesktop bug tracker.

[Bastien Nocera (hadess-deactivatedaccount)]

(In reply to comment #5)
> Hm. After reading some more about the formats, it should also be possible to
> extend the graphics a bit to be
>
> P1\x0a# (offset: 0)
>
> so the magic includes the (possible) comment in line 2, although this comment
> is not required. But it is less generic than just searching for P1..7.

Seems like the best option. A magic of 2 characters is just too short...

* freedesktop.org.xml.in: Tighten the magic for PBM,
PPM and PGM types, so as to avoid false positives
(Closes: #2359)
* tests/list:
* tests/test.pbm: Add a pbm test file

[Sebastien Bacher (seb128)]

the bug has been fixed upstream now

[Sebastien Bacher (seb128)]

the bug is fixed in the intrepid version

[Nunojsilva (nunojsilva)]

Some image viewers, which rely on gtk+ to show images, are unable to show portable anymaps. gtk+ seems to rely on freedesktop.org.xml for file type detection, which fails for some anymaps.

Removing the lines about # solves this, and here "man 5 ppm" does not say a comment is needed (although it can exist).