[asterisk-dev] Possible error in "datalen" member of example slin8 and slin16 frames

[Gianluca Merlo]

Hello,

while working on integrating a new codec translator (from/to slin16) into
Asterisk I came across a fact which stimulated my curiosity.

Specifically, I am conducting the first tests on the new translator by
observing how it works in recalculating translations, and I see that while
translating from slin16 to the new codec the framein function is called
with ast_frames which are not what I expected. The anomaly in these frames
is that they are reported (by their "samples" member) to consist of 160
samples, but to be (by their "datalen" member) 640 bytes in size. What I
expected, given the fact that each sample is 16 bit in size, is that 160
samples would mean a frame "datalen" of 320 bytes.

Exploring this strange phenomenon, I suspect that it is due to a small
imprecision in include/asterisk/slin.h, caused by commits related to the
issue 15700 (https://issues.asterisk.org/jira/browse/ASTERISK-15700).
Precisely, it seems to me that in the definition of slin8 and slin16
example frames, the "datalen" member is set to twice the real size of the
data in the frame.
The value of the member was correct before the commits in the issue, which
IMHO meant simply to change frame data type as uint16_t for ARM
compatibility, and consequently adjust the "samples" member to reflect the
fact that with uint16_t units, the number of sample in the frame was the
same as the total length of the data array. The unintended change is
multiplying "datalen" by two, as "sizeof" already gives the currect size of
the data in the frame.

I apologize if my explanation is a bit verbose, and would like your
confirmation that my guesses are correct and safe in regards to issue 15700
before opening an issue for a correction.

Best regards

Gianluca Merlo
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.digium.com/pipermail/asterisk-dev/attachments/20130702/5777b789/attachment.htm>