[tlbuild] makejvf.test test failures

[Ken Brown]

Hi Hironobu,

On 7/23/2017 5:42 AM, Hironobu Yamashita wrote:
> Thank you Mojca,
> 
> 2017/07/23 16:17, Mojca Miklavec <mojca.miklavec.lists at gmail.com>:
>>
>>  From hex dump:
>>
>> Original:
>> 00000350: 9101 8a1f 9ff7 33c0 8121 6df2 0000 0087  ......3..!m.....
>> 00000360: 0000 2149 0007 b29e 9106 287e 9f08 cc40  ..!I......(~...@
>>
>> The file built on that machine:
>> 00000350: 9101 8a1f 9ff7 33c1 8121 6df2 0000 0087  ......3..!m.....
>> 00000360: 0000 2149 0007 b29e 9106 287e 9f08 cc3f  ..!I......(~...?
>>
>> So:
>> 0x357: c1 instead of c0
>> 0x36f: 3f instead of 40

My hexdump looks the same, so I guess you don't need my tmin10.vf after all.

> These bytes correspond to the last byte of right/down shift amount
> of character JIS 0x2148 and JIS 0x2149 calculated in write.c,
> so these differences might come from round-off error.
> Therefore, the test failure is not critical for practice.
> 
> I'm not sure this test failure can be solved by modifying sources.
> Maybe I should replace test files to avoid such a round-off error?

Makes sense.

Ken