[libdbi-devel] Problems in dbi_result_get_binary_copy_idx

[Markus H. <mar...@mh...>]

Jo=E3o Henrique Freitas writes:
 > I have a problem in the function dbi=5Fresult=5Fget=5Fbinary=5Fcopy.=

 >=20
 > That function call the dbi=5Fresult=5Fget=5Fbinary=5Fcopy=5Fidx.
 >=20
 > In the code dbi=5Fresult.c on line 1310 and 1311. The size variable
 > always is 0 or a low value and malloc allocates insufficient space i=
n
 > pointer to newblob. The result is in line 1316 , memcpy not works li=
ke
 > expected.
 >=20
 > And so I think that has a little bug in dbi=5Fresult=5Fget=5Fbinary=5F=
copy=5Fidx.
 >=20

Your analysis is correct. The problem is caused by the improper use of
the field index in the call to
dbi=5Fresult=5Fget=5Ffield=5Flength=5Fidx(). libdbi uses a 0-based inde=
x
internally, but when an API function is called, we need to use the
user-visible 1-based index instead. Therefore this call retrieved the
size of an adjacent column, causing the blob to get truncated.

I've checked in a fix (see the attached diff for reference). I've also
checked in an extended version of your test=5Fdbi.c patch which covers
all database engines and also fixes a few minor errors in your
patch. I was a little surprised to learn that one of the API functions
was not covered by a test. Thanks for getting this to my attention.

regards,
Markus