[MPlayer-dev-eng] [PATCH] Add fate test

[Reimar Döffinger]

On 9 Nov 2011, at 02:43, Diego Biurrun <diego at biurrun.de> wrote:

> On Sun, Nov 06, 2011 at 11:04:10AM +0100, Reimar Döffinger wrote:
>> a bit hackish, video only, some samples fail even when testing on a
>> single configuration.
> 
>> --- Makefile    (revision 34309)
>> +++ Makefile    (working copy)
>> @@ -926,6 +926,7 @@
>> 
>> clean:
>>    -$(MAKE) -C ffmpeg $@
>> +    -$(MAKE) -C tests clean
>>    -rm -f $(call ADD_ALL_DIRS,/*.o /*.a /*.ho /*~)
>>    -rm -f $(call ADD_ALL_EXESUFS,mplayer mencoder)
>> 
>> @@ -1087,12 +1088,14 @@
>> 
>> +fatetest: mplayer$(EXESUF)
>> +    $(MAKE) -C tests fatetest
>> 
>> 
>> -include $(DEP_FILES) $(DRIVER_DEP_FILES) $(TESTS_DEP_FILES) $(TOOLS_DEP_FILES) $(DHAHELPER_DEP_FILES)
>> 
>> .PHONY: all doxygen *install* *tools drivers dhahelper*
>> -.PHONY: checkheaders *clean tests check_checksums
>> +.PHONY: checkheaders *clean tests check_checksums fatetest
> 
> I designed the build system non-recursive and monolithic on purpose.
> 
> tests/Makefile should be part of the main Makefile.

Well, including it should be no problem, but I really don't want everything in one single Makefile, that one is already almost as "nice" to get an overview of as mplayer.c
I remember you complaining that people don't apply good programming principles to build systems, so this is a good point at applying the principle "independent stuff I different files" IMO.
There's also some other minor annoyance like that it is convenient to run the test script from the tests directory, that's a bit more difficult with non-recursive make.

> 
>> --- tests/faterun.sh    (revision 0)
>> +++ tests/faterun.sh    (revision 0)
>> @@ -0,0 +1,10 @@
>> +#!/bin/sh
>> +i=$1
>> +echo "running $i"
>> +mkdir -p res/$(dirname $i)
>> +touch res/$i.md5
>> +../mplayer -noconfig all -lavdopts threads=4:bitexact -really-quiet -noconsolecontrols -nosound -benchmark -vo md5sum:outfile=res/$i.md5 $FATE_SAMPLES/$i
>> +if ! diff -uw ref/$i.md5 res/$i.md5 ; then
>> +    mv res/$i.md5 res/$i.md5.bad
>> +    exit 1
>> +fi
> 
> "i" is not really a better variable name than "1".
> Also, a few empty lines could make this more readable.

I had a loop in there originally and forgot to clean it up.
I is still better since it's easier to change i=$1 to i=$2 for example than having to change it everywhere.
>