Re: [LTP] stack_size test case default value wrong?

[Mike F. <va...@ge...>]

On Monday 02 April 2007, Subrata Modak1 wrote:
> The Unix98 standard requires malloc() to set errno to ENOMEM upon failure.

that whole description of malloc, while informative, is irrelevant here ...=
=20
nowhere is errno being checked by the stack_space.c test.  the error is tha=
t=20
the bits calculation comes out to 0 and while glibc will go ahead and=20
allocate some memory for you anyways, other implementations will return NUL=
L=20
(which is saner imo and certainly valid according to POSIX).

so on any system (like uClibc) where malloc(0) returns NULL, stack_space wi=
ll=20
bail early claiming malloc() failed:
    bits =3D (char*)malloc((nchunks+7)/8);
    val_buf =3D (char*)(malloc(csize));
    zero_buf =3D (char*)(malloc(csize));

    if (bits =3D=3D 0 || val_buf =3D=3D 0 || zero_buf =3D=3D 0) {
        tst_resm(TFAIL, "\tmalloc failed, pid: %d\n",getpid());
        tst_exit();
    }
=2Dmike