Derek Price wrote:
> >But it doesn't do so if malloc() fails on non-POSIX systems (like mingw
> >or so). I think one should set errno = ENOMEM if malloc() fails.
>
> How about the attached patch?
+ errno = 0;
+ unaligned_ptr = malloc (size + pagesize - 1);
+ if (unaligned_ptr == NULL)
+ {
+ /* Failed malloc on some non-posix systems (e.g. mingw) fail to set
+ errno. */
+ if (!errno) errno = ENOMEM;
+ return NULL;
+ }
malloc() can (and usually does) cause system calls that can modify errno.
I.e. on an ISO C compliant implementation, errno can be anything after
malloc() returns. I.e. it is not meaningful.
Also, looking at the value of _POSIX_C_SOURCE to determine whether malloc()
will set errno or not is doomed, because even in mingw, people can
#define _POSIX_C_SOURCE 2
or similar.
So the only thing I consider possible is the appended patch.
Bruno
diff -c -3 -r1.4 pagealign_alloc.c
*** pagealign_alloc.c 3 Mar 2005 20:38:38 -0000 1.4
--- pagealign_alloc.c 4 Mar 2005 12:23:25 -0000
***************
*** 149,155 ****
size_t pagesize = getpagesize ();
void *unaligned_ptr = malloc (size + pagesize - 1);
if (unaligned_ptr == NULL)
! return NULL;
ret = (char *) unaligned_ptr
+ ((- (unsigned long) unaligned_ptr) & (pagesize - 1));
new_memnode (ret, unaligned_ptr);
--- 149,160 ----
size_t pagesize = getpagesize ();
void *unaligned_ptr = malloc (size + pagesize - 1);
if (unaligned_ptr == NULL)
! {
! /* Set errno. We don't know whether malloc already set errno: some
! implementations of malloc do, some don't. */
! errno = ENOMEM;
! return NULL;
! }
ret = (char *) unaligned_ptr
+ ((- (unsigned long) unaligned_ptr) & (pagesize - 1));
new_memnode (ret, unaligned_ptr);
_______________________________________________
Bug-cvs mailing list
Bug-cvs@gnu.org
http://lists.gnu.org/mailman/listinfo/bug-cvs
