You can't. Right now this is a limitation of the compiler. If it's a
common use case we can likely find a solution.
Adam
On Dec 2, 2008, at 10:25 PM, Dan Mick wrote:
> How does one make the compiler generate runnable code with such D
> input?
>
> Adam Leventhal wrote:
>> Hey Jim,
>> I was able to reduce this to a simpler case:
>> # dtrace -n BEGIN'{ foo[1, 1, 1, 1, 1, 1] = 1; }'
>> dtrace: failed to enable 'BEGIN': DIF program content is invalid
>> This failed because the D compiler generated invalid DIF in that it
>> used
>> too many virutal registers.
>> If you try one more key, the error is properly identified.
>> # dtrace -n BEGIN'{ foo[1, 1, 1, 1, 1, 1, 1] = 1; }'
>> dtrace: invalid probe specifier BEGIN{ foo[1, 1, 1, 1, 1, 1, 1] =
>> 1; }:
>> Insufficient registers to generate code
>> It looks like libdtrace is happily allocating a 9th register even
>> though there are only 8:
>> # dtrace -c '/usr/sbin/i86/dtrace -s /var/tmp/foo.d' -n 'pid
>> $target::dt_regset_alloc:return{ trace(arg1); }'
>> dtrace: description 'pid$target::dt_regset_alloc:return' matched 1
>> probe
>> dtrace: failed to compile script /var/tmp/foo.d: Insufficient
>> registers to
>> generate code
>> dtrace: pid 101053 has exited
>> CPU ID FUNCTION:NAME
>> 0 73111 dt_regset_alloc:return 0
>> 0 73111 dt_regset_alloc:return 1
>> 0 73111 dt_regset_alloc:return 2
>> 0 73111 dt_regset_alloc:return 3
>> 0 73111 dt_regset_alloc:return 4
>> 0 73111 dt_regset_alloc:return 5
>> 0 73111 dt_regset_alloc:return 6
>> 0 73111 dt_regset_alloc:return 7
>> 0 73111 dt_regset_alloc:return 8
>> 0 73111 dt_regset_alloc:return 4294967295
>> It looks like there's a simple fix in dt_regset_create() which is
>> not to add
>> one to the number of registers we've been told exist. I've filed
>> this bug
>> for you:
>> 6778459 libdtrace will allocate one more register than there exists
>> Adam
>> On Sun, Nov 30, 2008 at 08:36:50PM -0800, James Litchfield wrote:
>>> What's going on?
>>>
>>> # dtrace -s iotime_all.d 100
>>> dtrace: failed to enable 'iotime_all.d': DIF program content is
>>> invalid
>>>
>>> The errant script....
>>>
>>> #pragma D option quiet
>>>
>>> BEGIN
>>> {
>>> stime = timestamp;
>>> io_count = 0;
>>> }
>>>
>>> io:::start
>>> /args[2]->fi_pathname != "<none>"/
>>> {
>>> start[pid, args[2]->fi_pathname, args[0]->b_edev, args[0]-
>>> >b_blkno, args[0]->b_bcount] = timestamp;
>>> self->pid = pid;
>>> self->name = args[2]->fi_pathname;
>>> self->size = args[0]->b_bcount;
>>> }
>>> io:::start
>>> /args[2]->fi_pathname != "<none>"/
>>> {
>>> start[pid, args[1]->dev_pathname, args[0]->b_edev, args[0]-
>>> >b_blkno, args[0]->b_bcount] = timestamp;
>>> self->pid = pid;
>>> self->name = args[1]->dev_pathname;
>>> self->size = args[0]->b_bcount;
>>> }
>>>
>>> io:::done
>>> /start[self->pid, self->name, args[0]->b_edev, args[0]->b_blkno,
>>> self->size]/
>>> {
>>> this->elapsed = timestamp - start[self->pid, self->name,
>>> args[0]->b_edev, args[0]->b_blkno, self->size];
>>> printf("%5u %10s %58s %2s %8u %8u %3d.%03d\n", self->pid,
>>> args[1]->dev_statname,
>>> self->name, args[0]->b_flags & B_READ ? "R" : "W",
>>> args[0]->b_bcount, self->size,
>>> this->elapsed / 1000000, (this->elapsed / 1000) % 1000);
>>> start[self->pid, self->name, args[0]->b_edev, args[0]->b_blkno,
>>> self->size] = 0;
>>> self->pid = 0;
>>> self->name = 0;
>>> self->size = 0;
>>> io_count++;
>>> }
>>> io:::done
>>> /io_count > $1/
>>> {
>>> printf("Elapsed Time: %u seconds\n\n", (timestamp - stime) /
>>> 1000000000);
>>> exit (0);
>>> }
>>>
>>> _______________________________________________
>>> dtrace-discuss mailing list
>>> dtrace-discuss@opensolaris.org
>
--
Adam Leventhal, Fishworks http://blogs.sun.com/ahl
_______________________________________________
dtrace-discuss mailing list
dtrace-discuss@opensolaris.org
