audiofile 0.2.7 fails to build (on testsuite)

[alessio]

As you know, I'm working on the update of the Debian packaging of audiofile 0.2.7.

While I've been testbuilding the new package, I've encountered a build failure when running the testsuite:

testfloat: testing AIFF-C
testfloat: testing WAVE
testfloat: testing NeXT .snd
testfloat: testing IRCAM
testfloat passed
PASS: testfloat
testdouble: testing AIFF-C
testdouble: testing MS RIFF WAVE
testdouble: testing NeXT .snd/Sun .au
testdouble passed
PASS: testdouble
PASS: testmarkers
PASS: testchannelmatrix
PASS: seek
Testing AIFF
Testing AIFF-C
Testing NeXT .snd/Sun .au
Incorrect frame count in file opened for reading

FAIL: large

#1 of 25 tests failed

Here is the full buildlog:
http://paste.debian.net/126367/

Would you help me to fix that, please?

[mpruett]

Unmodified Audio File Library version 0.2.7 passes the test suite on x86_64. Since 10_au_length_unspecified.dpatch in Debian's audiofile 0.2.6-8 package doesn't apply cleanly against version 0.2.7, it appears that patch is the source of this failure.

The patch below against version 0.2.7 is based upon commit 085132b. With this change running the test suite is successful.

diff -ru audiofile-0.2.7/libaudiofile/next.c audiofile-0.2.7-patched/libaudiofile/next.c
--- audiofile-0.2.7/libaudiofile/next.c 2010-03-18 20:03:58.000000000 -0700
+++ audiofile-0.2.7-patched/libaudiofile/next.c 2011-08-16 11:48:12.163183373 -0700
@@ -64,6 +64,8 @@
    NULL            /* miscellaneous */
 };

+static const uint32_t _AU_LENGTH_UNSPECIFIED = 0xffffffff;
+
 status _af_next_read_init (AFfilesetup setup, AFfilehandle file)
 {
    uint32_t    id, offset, length, encoding, sampleRate, channelCount;
@@ -111,7 +113,11 @@
    track->f.compressionType = AF_COMPRESSION_NONE;

    track->fpos_first_frame = offset;
-   track->data_size = af_flength(file->fh) - offset;
+
+   off_t lengthAvailable = af_flength(file->fh) - offset;
+   if (length == _AU_LENGTH_UNSPECIFIED || length > lengthAvailable)
+       length = lengthAvailable;
+   track->data_size = length;

    switch (encoding)
    {

[alessio]

OK, so now the new 10_au_length_unspecified.patch looks so: http://paste.debian.net/126485/
What do you think about it?

[mpruett]

The previous version of 10_au_length_unspecified.dpatch is unnecessary, so what I posted above is the entirety of the patch:

http://paste.debian.net/126525/