Dear gnulib developers,
In the manual (safe-alloc.texi), I read for the definition of
REALLOC_N:
If reallocation fails, the @code{ptr} variable is not modified.
This claim is justified a little bit earlier in the text:
Some of the memory allocation mistakes that are commonly made are
@itemize @bullet
…
@item
leaking memory in calls to @code{realloc} when that call fails.
However, the definition in safe-alloc.h does not repeat this claim in
the docstring. The macro definition is:
#define REALLOC_N(ptr, count) \
safe_alloc_check ((ptr) = reallocarray (ptr, count, sizeof *(ptr)))
I think it means that ptr is always replaced, and is set to NULL if the
reallocation failed. Do I interpret this definition correctly? If yes,
then it means that the old array leaks when realloc failed.
Is this a bug, or did I misunderstand something in the safe-alloc
module?
Best regards,
Vivien
