Do we care about malloc failure in tests?

[Paul Pluzhnikov]

Greetings,

We have a few bugs opened by Bill, e.g.: BZ #19672, BZ #19668
complaining about this pattern in tests:

  char *p = malloc(20);
  p[0] = '\0';  // did not check against NULL.

I spot-checked other existing tests:

libio/tst-atime.c
libio/tst-fopenloc.c
libio/tst-ftell-partial-wide.c
libio/tst-mmap2-eofsync.c

Two of them check for NULL, and two do not.

Do we have a policy here?
Should I fix them to consistently check for NULL, or delete existing
NULL checks?

Thanks,
-- 
Paul Pluzhnikov