IMGENC1_control() usage

Alexander Stohr

as i am having some code with the JPEG encoder (not the decoder) based on "imgenc1.h" at my hands i selected to test a little bit with it.

look at this snippet:

Juan Gonzales said:
status = IMGDEC_control( hDec, XDM_RESET, &idecDynParams, &idecStatus );

f( status != IMGDEC_EOK ){
printf( "IMGDEC_control(XDM_RESET): error1\n" );
break;
}

sprufe5c says this:

XDM_GETBUFINFO: get required number of I/O buffers and their sizes. Results are
returned in the bufInfo member of the structure IIMGENC1_Status whose pointer is
passed as 4th argument.
The third argument is a pointer to a dynamic params structure of type
IIMGENC1_DynamicParams or IJPEGENC_DynamicParams (typecast to the previous
one). This argument is used whenever command ID is XDM_SETPARAMS.

so i thought for the other commands the third parameter is dont care and thus can even be set to NULL.

bad guess - the call for XDM_GETBUFINFO resulted in a segmentation fault from inside CE, in a function called JPEGENC_TI_control as reported by gdb.

so the codec does not require it at all but the embedding codec engine is touching it for some reasons.

the return value naming might as well be calling level dependent:

* IALG_OK/IALG_FAIL/other for the codec API

* IMGENC_* or IMGENC1_* for the VISA API subset.

over 15 years ago

0 Chris Ring over 15 years ago

TI__Genius 17205 points

The XDM spec for the IIMGENC1_Fxns.control() fxn states that the 3rd and 4th arguments are mandatory.

params must not be NULL, and must point to a valid IIMGENC1_DynamicParams structure.
status must not be NULL, and must point to a valid IIMGENC1_Status structure.

The CE IMGENC1_control() API guide could make this clearer, and I'll add that to the docs... but in any event, the XDM spec states these aren't optional args. So providing NULL is an API violation.

An insignificant detail - the JPEGENC_TI_control() fxn is actually a codec fxn, not a part of the Codec Engine framework. So it's actually the codec that's touching this struct (as the XDM spec says it should be able to).

Another tip, if you are using a recent enough CE release, you could use the CE_CHECK feature - which would check for this spec violation (and others) and return an error, rather than blindly pass the NULL param to the codec.

Chris

0 Alexander Stohr over 15 years ago in reply to Chris Ring

Expert 1200 points

i qu0ted sprufe5c in above statement. this is the "Sequential JPEG Encoder Codec on DM355" users guide.

so the fault comes from the codec but its documentation is a bit short about the parameter requirements.

0 Yashwant Dutt over 15 years ago in reply to Alexander Stohr

TI__Genius 17215 points

Agreed, codec documentation can be more clearer. Will make it more elaborate whenever we plan the next release.

regards

Yashwant

Processors

Processors forum

IMGENC1_control() usage