"encode_decode_test.c" infinite loop while encoding

[GoogleCodeExporter]

What steps will reproduce the problem?

1. Use the attached "test_source" and "test_input" files as the source and
input for an encode operation using the encode_decode_test.c

2. Watch the code lock up infinite loop.

3. After forcefully closing the application, a delta file will be produced.
I've attached the delta_output I received. I'm not sure whether this delta
is valid or not, but when running the decoding process using the
"test_source" and the "delta_output", the file it produces does not equal
"test_input". 


Using xdelta3.0t on Windows Vista, using Visual C++ 2008 (VC9).

Code example, just replace the appropriate lines in the "encode_decode_test.c"

Code for step 1:

    InFile = fopen("C:\\test_input", "rb");
    SrcFile = fopen("C:\\test_source", "rb");
    OutFile = fopen("C:\\delta_output", "wb");

    r = code (1, InFile, SrcFile, OutFile, 0x1000);

    fclose(OutFile);
    fclose(SrcFile);
    fclose(InFile);

Code for step 2:

    InFile = fopen("C:\\delta_output", "rb");
    SrcFile = fopen("C:\\test_source", "rb");
    OutFile = fopen("C:\\result.abc", "wb");

    r = code (0, InFile, SrcFile, OutFile, 0x1000);

    fclose(OutFile);
    fclose(SrcFile);
    fclose(InFile);

Original issue reported on code.google.com by wyatto...@gmail.com on 12 Apr 2008 at 7:39

Attachments:

• test_input
• test_source
• delta_output

[GoogleCodeExporter]

Thank you!

This was an excellent test case.  The problem was fairly easy to discover, and 
shows
(for the second time) that I need to add the non-blocking encode_decode_test to 
my
comprehensive test suite.

I'd like to hold off on making a new release for a little while, but the 
problem has
been fixed in SVN 234.

I've made some compilation fixes and added command-line arguments to
encode_decode_test, to make testing easier.

Original comment by josh.mac...@gmail.com on 15 Apr 2008 at 4:03

• Changed state: Fixed

[GoogleCodeExporter]

A more proper fix is in SVN 239.  This is a really great test case because it's 
hard
to construct by hand (requires hash collisions).

Original comment by josh.mac...@gmail.com on 22 Apr 2008 at 2:45

[GoogleCodeExporter]

I've been working on a new regression test and just added a test for this case.

Original comment by josh.mac...@gmail.com on 10 Jul 2008 at 4:09