LIBFML_CAT:1: ERROR: Fielded buffer not aligned

Question

Hi,
Can anyone help me with this error.
LIBFML_CAT:1: ERROR: Fielded buffer not aligned
I receive this error while initializing a FML32 buffer. Following is the code snippet
tempBuf=(FBFR32 *) tpalloc("FML32",NULL,(FLDLEN32)Fneeded32(19, sizeof(vi_t)));
if (tempBuf == NULL)
printf("\nError allocating tempBuf<%s>\n", tpstrerror(tperrno));
exit(1);
if (Finit32(tempBuf, (FLDLEN32)Fneeded32(19, Fsizeof(tempBuf)) == -1))
printf("\nError initializing tempBuf<%s>\n", Fstrerror32(Ferror32));
exit(1);
Thanks.

Answer 1

Hi,
Here are some observations on this code snippet:
1) There are parentheses missing on the Finit32() line so the second argument to Finit32() is
(FLDLEN32)Fneeded32(19, Fsizeof(tempBuf)) == -1 .
Thus, the second argument will be 0 if Fneeded32() succeeds and will be 1 if Fneeded32() returns -1.
An Finit32() buflen argument of 0 will result in a -1 return with Ferror32 set to FNOSPACE and an Finit32() buflen argument of 1 will result in a -1 return with Ferror32 set to FALIGNERR (since 1 is not a multiple of 4.) This is why you are seeing the FALIGNERR problem.
2) This code is using 32-bit FML, but Fsizeof() is a 16-bit FML function (assuming that you have not included <fml1632.h>.) Since tempBuf is an FML32 buffer, attempts to use 16-bit FML on this buffer will return -1, which means that a -1 length will be pased to Fneeded32, which will return -1, and -1 == -1, so a second argument of 1 will be passed to Finit32.
3) Changing Fsizeof(tempBuf) to Fsizeof32(tempBuf) would still result in code with problems, since Fsizeof(tempBuf) is greater than sizeof(vi_t). Thus, the seond Fneeded32() will return a bigger size than the first Fneeded32(), which will result in initializing memory past the end of the buffer. As stated on the Finit32() manual page (modified for 32-bit FML) "The correct way to reinitialize a buffer to have no fields is: Finit32(fbfr, (FLDLEN32)Fsizeof32(fbfr));".
4) It should not be necessary to call Finit32() immediately after tpalloc() since tpalloc() will have initialized the FML32 buffer.
Regards,
Ed