Re: [PATCH] IDE-TAPE NULL terminate strings.

[Mark de Wever <koraq@xs4all.nl>]

On Tue, Sep 23, 2008 at 09:10:43PM +0400, Sergei Shtylyov wrote:
> Mark de Wever wrote:
>
>>>> diff --git a/drivers/ide/ide-tape.c b/drivers/ide/ide-tape.c
>>>> index 1bce84b..c41f5b1 100644
>>>> --- a/drivers/ide/ide-tape.c
>>>> +++ b/drivers/ide/ide-tape.c
>>>> @@ -2338,7 +2338,7 @@ static void idetape_get_inquiry_results(ide_drive_t *drive)
>>>> {
>>>> 	idetape_tape_t *tape = drive->driver_data;
>>>> 	struct ide_atapi_pc pc;
>>>> -	char fw_rev[6], vendor_id[10], product_id[18];
>>>> +	char fw_rev[6] = {'\0'}, vendor_id[10] = {'\0'}, product_id[18] = {'\0'};
>
>>>  Do you realize how much *absolutely unnecessary* code will this bring
>>> in?
>
>> I did not, I just had a look at the code GCC produced. I did expect much
>> smaller code, but maybe that's only generated with -Os.
>
>    My imagination sufficed to foresee how much code a compiler would have 
> to produce to completely initialize the arrays of the *auto* memory class 
> -- even regardless of optimization. And all that mostly to no purpose.

I expected a rep stos based code, which isn't much code. But the code
was created without a rep, instead it used several stos opcodes. I agree
with the mostly no purpose part, since the memcpy later overwrites most
of the array.

>>> This is certainly worse than your initial patch (if it was correct).
>
>> My initial patch did work, 
>
>    If ide_fixstring() wouldn't have to do any space compression, it would 
> work. If it would have to compress spaces, 2 garbage characters would be  
> copied by it and then printed.

I didn't know that, I guess I got lucky ;-)

>> but that doesn't matter much, since Boris
>> posted another patch based on your suggestions. I like that patch better
>> as my initial patch. I'm testing it now and I expect it to work.
>
>    Me too. :-)

It worked, thanks for your suggestion :-)

Regards,
Mark de Wever