Re: [PATCH v9 0/7] coroutines: generate wrapper code

[Eric Blake <eblake@redhat.com>]

On 9/25/20 3:04 AM, Vladimir Sementsov-Ogievskiy wrote:
> 24.09.2020 23:32, no-reply@patchew.org wrote:
>> Patchew URL: 
>> https://patchew.org/QEMU/20200924185414.28642-1-vsementsov@virtuozzo.com/
>>

>> Program python3 found: YES (/usr/bin/python3)
>> Configuring ninjatool using configuration
>> ---
>>      return codecs.ascii_decode(input, self.errors)[0]
>> UnicodeDecodeError: 'ascii' codec can't decode byte 0xe2 in position 
>> 11406: ordinal not in range(128)

> Generating 'libqemu-aarch64-softmmu.fa.p/decode-vfp.c.inc'.
> Traceback (most recent call last):
>    File 
> "/tmp/qemu-test/src/block/../scripts/block-coroutine-wrapper.py", line 
> 187, in <module>
>      f_out.write(gen_wrappers(f_in.read()))
>    File "/usr/lib64/python3.6/encodings/ascii.py", line 26, in decode
>      return codecs.ascii_decode(input, self.errors)[0]
> UnicodeDecodeError: 'ascii' codec can't decode byte 0xe2 in position 
> 11406: ordinal not in range(128)
> 
> 
> Interesting:
> 
> [root@kvm up-coroutine-wrapper]# grep --color='auto' -P -n 
> '[^\x00-\x7F]' include/block/block.h
> 307:     * Child from which to read all data that isn’t allocated in the
>                                                       ^
> 
> The file really contains one non-ascii symbol. I think it worth a 
> separate patch. Still, it shouldn't break build process. On my system it 
> works as is, probably unicode is default for me.

Python 3 has had an interesting history when it comes to 8-bit cleanness 
by default.  Which means we DO have to be explicit about utf8.

> 
> Aha, from "open" specification:
> 
>     if encoding is not specified the encoding used is platform 
> dependent: locale.getpreferredencoding(False) is called to get the 
> current locale encoding.
> 
> 
> 
> Is it ok, that utf-8 is not default on test system?

It's intentional.

> 
> So, possible solutions are:
> 
> 1. Enforce utf-8 io in scripts/block-coroutine-wrapper.py (patch 4)

Yes, we should do that regardless (we do it in our other python scripts).

> 2. Drop non-ascii quotation mark from block.h

Yes, we should do that as well (it's only in a comment, but it is 
inconsistent).

> 3. Fix the test system default to be utf-8

No.  That one we want to keep where it is, because it helps us flush out 
these sorts of issues.

> 
> Do we want them all?
> 

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