the current shared-mime-info (0.17-0ubuntu7) contains 3 possible matches (all at
the same priority, 50) for files starting with "%"
this causes PDF files to have their typed detected as octet-stream.
The TeX/matlab matches should have their priority lowered so that the more
specific "%PDF-" string will match with higher priority.
TeX and matlab having the same magic detection string seems vaguely useless,
Note: postscript starts like "%!PS-Adobe-2.0" and is similarly broken by this bug.
I can confirm this with shared-mime-info-0.17-1.fc5.1,
on Fedora Core 5. Some PDFs are correctly identified,
others as text. I increased the priority for PDF from 50 to 60,
now all PDFs are correctly identified by nautilus.
(In reply to comment #2)
> I can confirm this with shared-mime-info-0.17-1.fc5.1,
> on Fedora Core 5. Some PDFs are correctly identified,
> others as text. I increased the priority for PDF from 50 to 60,
> now all PDFs are correctly identified by nautilus.
Could you please tell me where to increase the priority, I have some serious
troubles sending pdf which are not properly identified via email.
Confirm on Debian unstable (shared-mime-info 0.17-1 package). Changing the glob
priority of PDF from 50 to 60 fixes the problem. Any progress on getting this
actually noticed by someone with commit privs?
Here's the downstream report on Fedora Core 5:
Interestingly enough 0.17-1.fc5.1 was supposed to fix a postscript vs. matlab
* Wed Mar 22 2006 Matthias Clasen <email@example.com> - 0.17-1.fc5.1
- Backport upstream change to fix postscript vs. matlab confusion
The issue is also solved by upgrading "shared-mime-info" to current "CVS",
as I had pointed out in:
The "RPM" package built accordingly on 05/25/06 behaves correctly.
(In reply to comment #6)
> The issue is also solved by upgrading "shared-mime-info" to current "CVS",
> as I had pointed out in:
Since it looks like it's been fixed in freedesktop's CVS, by this change:
it should be marked closed.