Bug#904178: ffmpeg: fix mips r6 build

2018-07-21 Thread James Cowgill
Control: notfound -1 4.0.1-2
Control: found -1 7:4.0.1-2

Hi Yunqiang,

On 21/07/18 05:46, YunQiang Su wrote:
> Package: src:ffmpeg
> Version: 4.0.1-2
> 
> We are working on MIPS r6 ports.
> 
> Please change the lines about mips options in debian/rules to the below.
> # Disable some optimizations on mips*, because they are not always
> available on these architectures
> ifneq (,$(filter mips%r6 mips%r6el, $(DEB_HOST_ARCH)))
> MIPS_NO_R6_OPTION = --disable-mips32r2 \
> --disable-mips64r2 \
> --disable-mips32r5

Is this line intentional (given that --disable-mips32r5 is used below as
well)?

I don't think the name of that variable explains itself very well,
especially since one of the options enables R6. Maybe MIPS_R2_R6_CONFIG
is better?

James



signature.asc
Description: OpenPGP digital signature


Bug#904178: ffmpeg: fix mips r6 build

2018-07-21 Thread YunQiang Su
James Cowgill  于2018年7月21日周六 下午11:43写道:
>
> Hi,
>
> On 21/07/18 16:38, YunQiang Su wrote:
> > James Cowgill  于2018年7月21日周六 下午11:16写道:
> >> On 21/07/18 05:46, YunQiang Su wrote:
> >>> Please change the lines about mips options in debian/rules to the below.
> >>> # Disable some optimizations on mips*, because they are not always
> >>> available on these architectures
> >>> ifneq (,$(filter mips%r6 mips%r6el, $(DEB_HOST_ARCH)))
> >>> MIPS_NO_R6_OPTION = --disable-mips32r2 \
> >>> --disable-mips64r2 \
> >>> --disable-mips32r5
> >>
> >> Is this line intentional (given that --disable-mips32r5 is used below as
> >> well)?
> >>
> >
> > yes. when the first try, I don't add --disable-mips32r5, then it
> > failed to build,
> > when link r2/r6 objects.
>
> I still don't understand. In your new code, the --disable-mips32r5
> option is passed unconditionally in the main config block so this extra
> option in the r6 block should have no effect.

Ohh. you are right.
The current version disable mips32r5 unconditionally, and I pass it
twice for r6.
So, just remove that line in the condition of r6 will be OK.

ifneq (,$(filter mips%r6 mips%r6el, $(DEB_HOST_ARCH)))
MIPS_NO_R6_OPTION = --disable-mips32r2 \
--disable-mips64r2
else
MIPS_NO_R6_OPTION = --disable-mips32r6 \
--disable-mips64r6
endif

ifneq (,$(filter mips%, $(DEB_HOST_ARCH)))
CONFIG += --disable-mipsdsp \
--disable-mipsdspr2 \
--disable-loongson2 \
--disable-loongson3 \
--disable-mmi \
--disable-mips32r5 \
$(MIPS_NO_R6_OPTION) \
--disable-msa
endif



>
> >> I don't think the name of that variable explains itself very well,
> >> especially since one of the options enables R6. Maybe MIPS_R2_R6_CONFIG
> >> is better?
>
> James
>


-- 
YunQiang Su



Bug#904178: ffmpeg: fix mips r6 build

2018-07-21 Thread James Cowgill
Hi,

On 21/07/18 16:38, YunQiang Su wrote:
> James Cowgill  于2018年7月21日周六 下午11:16写道:
>> On 21/07/18 05:46, YunQiang Su wrote:
>>> Please change the lines about mips options in debian/rules to the below.
>>> # Disable some optimizations on mips*, because they are not always
>>> available on these architectures
>>> ifneq (,$(filter mips%r6 mips%r6el, $(DEB_HOST_ARCH)))
>>> MIPS_NO_R6_OPTION = --disable-mips32r2 \
>>> --disable-mips64r2 \
>>> --disable-mips32r5
>>
>> Is this line intentional (given that --disable-mips32r5 is used below as
>> well)?
>>
> 
> yes. when the first try, I don't add --disable-mips32r5, then it
> failed to build,
> when link r2/r6 objects.

I still don't understand. In your new code, the --disable-mips32r5
option is passed unconditionally in the main config block so this extra
option in the r6 block should have no effect.

>> I don't think the name of that variable explains itself very well,
>> especially since one of the options enables R6. Maybe MIPS_R2_R6_CONFIG
>> is better?

James



signature.asc
Description: OpenPGP digital signature


Bug#904178: ffmpeg: fix mips r6 build

2018-07-21 Thread YunQiang Su
James Cowgill  于2018年7月21日周六 下午11:16写道:
>
> Control: notfound -1 4.0.1-2
> Control: found -1 7:4.0.1-2
>
> Hi Yunqiang,
>
> On 21/07/18 05:46, YunQiang Su wrote:
> > Package: src:ffmpeg
> > Version: 4.0.1-2
> >
> > We are working on MIPS r6 ports.
> >
> > Please change the lines about mips options in debian/rules to the below.
> > # Disable some optimizations on mips*, because they are not always
> > available on these architectures
> > ifneq (,$(filter mips%r6 mips%r6el, $(DEB_HOST_ARCH)))
> > MIPS_NO_R6_OPTION = --disable-mips32r2 \
> > --disable-mips64r2 \
> > --disable-mips32r5
>
> Is this line intentional (given that --disable-mips32r5 is used below as
> well)?
>

yes. when the first try, I don't add --disable-mips32r5, then it
failed to build,
when link r2/r6 objects.

> I don't think the name of that variable explains itself very well,
> especially since one of the options enables R6. Maybe MIPS_R2_R6_CONFIG
> is better?
>
> James
>


-- 
YunQiang Su



Bug#904178: ffmpeg: fix mips r6 build

2018-07-20 Thread YunQiang Su
Package: src:ffmpeg
Version: 4.0.1-2

We are working on MIPS r6 ports.

Please change the lines about mips options in debian/rules to the below.
# Disable some optimizations on mips*, because they are not always
available on these architectures
ifneq (,$(filter mips%r6 mips%r6el, $(DEB_HOST_ARCH)))
MIPS_NO_R6_OPTION = --disable-mips32r2 \
--disable-mips64r2 \
--disable-mips32r5
else
MIPS_NO_R6_OPTION = --disable-mips32r6 \
--disable-mips64r6
endif

ifneq (,$(filter mips%, $(DEB_HOST_ARCH)))
CONFIG += --disable-mipsdsp \
--disable-mipsdspr2 \
--disable-loongson2 \
--disable-loongson3 \
--disable-mmi \
--disable-mips32r5 \
$(MIPS_NO_R6_OPTION) \
--disable-msa
endif

-- 
YunQiang Su