Simple text files interpreted as mtree

[kwrobot]

Original issue 399 created by Google Code user JanWielemaker on 2015-02-04T15:02:30.000Z:

<b>What steps will reproduce the problem?</b>
1. Create a file with content this content (a single line without quotes):
  &quot;b704c67b39978aff8cc254e78976f001ea07a6d0&quot;
2. Open the archive using format all|raw
3. Run archive_read_next_header() on ther archive

<b>What is the expected output? What do you see instead?</b>

Expected:
Get the plain entry, reported as &quot;data&quot;, format &quot;raw&quot;

Instead:
ERROR: Missing type keyword in mtree specification

<b>What version are you using?</b>

3.1.2

<b>On what operating system?</b>

Ubuntu

<b>How did you build?  (cmake, configure, or pre-packaged binary)</b>

pre-packaged

<b>What compiler or development environment (please include version)?</b>


<b>Please provide any additional information below.</b>

According to mtree(5), mtree files must start with
&quot;#mtree&quot;.  How can it be that the file is misinterpreted?

[kwrobot]

Comment #1 originally posted by kientzle on 2015-02-05T05:00:49.000Z:

mtree(5) is wrong.  There is no required header text.

That said, it should be possible to improve the mtree bidder to reject files that are obviously missing required text.

Mixing raw + mtree is a particularly risky combination, since it is very hard to distinguish mtree from plain text files.   I generally recommend that people not mix raw with any other format handler except 'empty.'

[kwrobot]

Comment #2 originally posted by kientzle on 2015-02-22T20:16:14.000Z:

I've started to put together a test that tries to verify that the mtree bidder is doing the right thing.

Please take a look at the attached file and see if these cases all make sense.  If you can think of other cases that would help us to fine-tune the mtree bid logic, please let us know.

Note:  The current code gets most of these right, but not all.

See attachment: test_read_format_mtree_bid.c