Re: [PATCH RFC] video: fbdev: sis: condition with no effect

[Thomas Winischhofer]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Tormod Volden wrote:
> On Thu, Feb 5, 2015 at 9:45 PM, Scot Doyle wrote:
>> On Wed, 4 Feb 2015, Nicholas Mc Guire wrote:
>>> The if and the else branch code are identical - so the condition has no
>>> effect on the effective code - this patch removes the condition and the
>>> duplicated code.
>>>
>>> Signed-off-by: Nicholas Mc Guire <hofrat@xxxxxxxxx>
>>> ---
>>>
>>> This code has been in here since commit 544393fe584d ("sisfb update") so I guess it is
>>> safe to simply remove the duplicated code if nobody noticed for 10 years.
>>>
>>> Note that the code is not really CodingStyle compliant - the lines inserted were formatted
>>> to satisfy the coding style but I'm unsure if it is not better to leave it in the
>>> old format.
>>>
>>> Patch was only compile tested with x86_64_defconfig +
>>> CONFIG_FB_SIS=m, CONFIG_FB_SIS_300=y, CONFIG_FB_SIS_315=y
>>>
>>> Patch is against 3.19.0-rc7 (localversion-next is -next-20150204)
>>>
>>>  drivers/video/fbdev/sis/init301.c |    9 ++-------
>>>  1 file changed, 2 insertions(+), 7 deletions(-)
>>>
>>> diff --git a/drivers/video/fbdev/sis/init301.c b/drivers/video/fbdev/sis/init301.c
>>> index 295e0de..9533a8ab 100644
>>> --- a/drivers/video/fbdev/sis/init301.c
>>> +++ b/drivers/video/fbdev/sis/init301.c
>>> @@ -7971,13 +7971,8 @@ SiS_SetCHTVReg(struct SiS_Private *SiS_Pr, unsigned short ModeNo, unsigned short
>>>           }
>>>        } else {                                               /* ---- PAL ---- */
>>>           /* We don't play around with FSCI in PAL mode */
>>> -         if(resindex == 0x04) {
>>> -            SiS_SetCH70xxANDOR(SiS_Pr,0x20,0x00,0xEF);       /* loop filter off */
>>> -            SiS_SetCH70xxANDOR(SiS_Pr,0x21,0x01,0xFE);       /* ACIV on */
>>> -         } else {
>>> -            SiS_SetCH70xxANDOR(SiS_Pr,0x20,0x00,0xEF);       /* loop filter off */
>>> -            SiS_SetCH70xxANDOR(SiS_Pr,0x21,0x01,0xFE);       /* ACIV on */
>>> -         }
>>> +       SiS_SetCH70xxANDOR(SiS_Pr, 0x20, 0x00, 0xEF); /* loop filter off */
>>> +       SiS_SetCH70xxANDOR(SiS_Pr, 0x21, 0x01, 0xFE); /* ACIV on */
>>>        }
>>>
>>>  #endif  /* 300 */
>> The code covering the PAL case had this redundancy when it was introduced
>> in Linux 2.4.19.
>>
>> Lines 7934-7981 consider three variables: PAL, overscan, and resindex.
>> Given the "#ifdef 0" block, couldn't the current six sections collapse
>> into two? One for (!PAL && overscan && resindex==5) and another for the
>> rest?
> 
> Are we sure there isn't a typo in one of the duplicate clauses? Or
> wrong copy-pasting? Generally I am skeptical to "fixing" code without
> understanding what is behind or testing it, and just cosmetically
> brush over it. For now at least it is obvious that there is something
> wrong. In case (although an unlikely one) someone who understands the
> code and knows this chip comes along, he would quickly spot this.
> After your "fixups" this will be all forgotten. Additionally it adds
> to the impression that this code is being maintained, which is wrong.
> 
> I would understand an argument about annoying compiler warnings and
> the like, but in that case I would prefer to #if 0 it instead of
> "prettifying" it.
> 
> 0.02
> Tormod
> 

The code is partly unfinished due to a lack of hardware to test this
with. SiS announced SiS+Chrontel 7019 combos at some point but I have
never seen one. The code was written based on the Chrontel datasheets,
which weren't clear to some extent, and there wasn't ever any test
hardware. I don't recall this one exactly, but identical if-else
statements mean that one alternative is (assumingly) correct, while the
other is uncertain and/or untested. I left such redunant if-statements
in the code to remember the conditions and the fact that there is a
second alternative.

Considering the long time I'd say it's safe to simplify this.

A word on other changes I monitored recently: Please bear in mind that
with video hardware reading and writing registers is not simple like
reading and writing to memory. Sometimes reading causes an effect in the
hardware as well (latches, etc), so removing seemingly redundant
GetReg/SetReg sequences might actually have an effect.

Regards
Thomas


- --
Thomas Winischhofer
thomas AT winischhofer DOT net
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.7 (Darwin)

iD8DBQFU1GaUzydIRAktyUcRAuQlAJ9NL3moUDf0yUMbE9qi4L26hT69NwCcDOk2
GyZjN8fic9bITTtdK9OG0R8=
=s9pA
-----END PGP SIGNATURE-----
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/