[cairo] Implementing cairo_read_func_t/cairo_write_func_t properly

[Behdad Esfahbod]

On 16-01-01 01:01 PM, Bernhard Fischer wrote:
> On Friday 01 January 2016 12:14:03 Behdad Esfahbod wrote:
>> On 16-01-01 09:32 AM, Bernhard Fischer wrote:
>>> On Thursday 31 December 2015 09:54:39 Lawrence D'Oliveiro wrote:
>>>>> On Wed, 30 Dec 2015 15:27:13 +0100, Bernhard Fischer wrote:
>>>>>>> when I would return STATUS_READ_ERROR, then the calling function
>>>>>>> cannot detect how many bytes then actually have been read except in
>>>>>>> the case that I read 1 byte after the other.
>>>>>
>>>>> Does it matter? In any case, the file is clearly corrupted and the
>>>>> image decoding has failed.
>>>>>
>>>>> At least, that is my conclusion...
>>>
>>> Yes, it matters!
>>> cairo_..._stream() are stream reading functions. This is that it does not
>>> know in advance how many bytes are available in the file.
>>>
>>> The read_func_t is called with a buffer of a programmer-chosen length
>>> (from
>>> within cairo_..._stream()), e.g. 10k. But if the file has just 9k,
>>> obviously only 9k are read. Unfortunately, the programmer has no way to
>>> find that out.
>> I agree this could be designed better...
> 
> I know that this is now getting deep into fundamental design issues.
> 
> I'd suggest a new prototype for that purpose. See the following file line 
> numbers 17 - 32. I experimentally implemented it into my jpeg stream reader 
> function (optionally chosen by cpp macro):
> 
> https://github.com/rahra/cairo_jpg/blob/master/src/cairo_jpg.h

You can have both at the same time.

b