Re: [PATCH v2 2/3] hw/mips/gt64xxx: Remove dynamic field width from trace events

[Eric Blake <eblake@redhat.com>]

On 11/18/19 1:10 PM, Philippe Mathieu-Daudé wrote:

>>>> -        trace_gt64120_write("INTRCAUSE", size << 1, val);
>>>> +        trace_gt64120_write("INTRCAUSE", size << 3, val);
>>>
>>> Again, this isn't mentioned in the commit message.  Why are you 
>>> changing parameter values?
>>>
>>>
>>>> +++ b/hw/mips/trace-events
>>>> @@ -1,4 +1,4 @@
>>>>   # gt64xxx.c
>>>> -gt64120_read(const char *regname, int width, uint64_t value) 
>>>> "gt64120 read %s value:0x%0*" PRIx64
>>>> -gt64120_write(const char *regname, int width, uint64_t value) 
>>>> "gt64120 write %s value:0x%0*" PRIx64
>>>> +gt64120_read(const char *regname, int width, uint64_t value) 
>>>> "gt64120 read %s width:%d value:0x%08" PRIx64
>>>> +gt64120_write(const char *regname, int width, uint64_t value) 
>>>> "gt64120 write %s width:%d value:0x%08" PRIx64
>>>
>>> Huh, we were really broken - the old code (if passed to printf) would 
>>> try to parse 4 parameters, even though it was only passed 3.  But it 
>>> looks like you still need a v3.
>>
>> Oops. I am surprise the compiler doesn't emit a warning here...
> 
> I'm sorry I can't see the 4th parameter.

My fault for chasing a red herring.  I guess I was mixing the three % 
post-patch with the two % and one * pre-patch, and somehow mis-counting 
three % plus one * as four arguments.  But re-reading your confusion, 
yes, there were only three parameters being consumed.

> 
> Before: "gt64120 read %s value:0x%0*" PRIx64
> 
> #1 's' for 'const char *regname'
> #2 '0*' for 'int width'
> #3 'x' for 'uint64_t value'
> 
> After: "gt64120 read %s width:%d value:0x%08" PRIx64
> 
> #1 's' for 'const char *regname'
> #2 'd' for 'int width'
> #3 '08x' for 'uint64_t value'
> 
> Am I missing something?

Rather, I was.

-- 
Eric Blake, Principal Software Engineer
Red Hat, Inc.           +1-919-301-3226
Virtualization:  qemu.org | libvirt.org