David Nicol <[EMAIL PROTECTED]> writes:
> On 5/12/05, Ton Hospel <[EMAIL PROTECTED]> wrote:
> > And isn't a simular fix needed for the returncode of system() ?
>
> fascinating stuff. Would it make sense to abstract return code massaging
> into a macro, possibly different per-platform?
It is already a macro and this macro is already used on the returncode
of system. The macro was the one that I patched (indirectly).
Another way to fix this to to make $? be exactly those 16 bits
described by this code segment found in description of system in
perlfunc:
| You can check all the failure possibilities by inspecting
| C<$?> like this:
|
| if ($? == -1) {
| print "failed to execute: $!\n";
| }
| elsif ($? & 127) {
| printf "child died with signal %d, %s coredump\n",
| ($? & 127), ($? & 128) ? 'with' : 'without';
| }
| else {
| printf "child exited with value %d\n", $? >> 8;
| }
and then we could drop this text:
| or more portably by using the W*() calls of the POSIX extension;
| see L<perlport> for more information.
and then introduce a variable like ${^NATIVE_CHILD_ERROR} that would
hold the real 'status' from wait*(2). The ${^NATIVE_CHILD_ERROR}
would be to $? what $^E is to $!.
Regards,
Gisle
