[cairo] Fwd: Re: [PATCH] Generate better error message when loading invalid PNGs

[Uli Schlachter]

Am 26.03.2016 um 13:17 schrieb Adrian Johnson:
> I forgot to CC the list.
> 
> -------- Forwarded Message --------
> Subject: Re: [cairo] [PATCH] Generate better error message when loading
> invalid PNGs
> Date: Sat, 26 Mar 2016 22:35:02 +1030
> From: Adrian Johnson <ajohnson at redneon.com>
> To: Uli Schlachter <psychon at znc.in>
> 
> On 26/03/16 18:05, Uli Schlachter wrote:
>> Am 25.03.2016 um 22:11 schrieb Adrian Johnson:
>>> On 26/03/16 03:53, Uli Schlachter wrote:
>>>> Based on an idea from Cyril Roelandt, this patch makes cairo generate better
>>>> error messages when cairo_image_surface_create_from_png{,_stream} is called on
>>>> e.g. JPEG files. To do so, we don't let libpng check if the file starts with a
>>>> PNG signature, but do so by hand. On mismatch, a surface with status
>>>> CAIRO_STATUS_READ_ERROR is returned where previously the status was
>>>> CAIRO_STATUS_NO_MEMORY (the status used for all errors in libpng).
>>>
>>> What about if it is a PNG file but libpng is unable to decode it? It is
>>> still going to return the unhelpful CAIRO_STATUS_NO_MEMORY. It is also
>>> useful restrict CAIRO_STATUS_READ_ERROR to errors reading the from the
>>> stream.
>>>
>>> I think we need an error status for backends that can fail. The attached
>>> patch adds CAIRO_STATUS_PNG_ERROR. I can also write a patch for win32 as
>>> it is returning CAIRO_STATUS_NO_MEMORY for GDI errors.
>>
>> Fine with me. Note that we could detect OOM reliably by making libpng use our
>> own malloc/free wrappers via png_create_{read,write}_struct_2(). However,
>> perhaps this is a bit excessive.
>>
>>> @@ -744,6 +746,7 @@ read_png (struct png_read_closure_t *png_closure)
>>>   *	%CAIRO_STATUS_NO_MEMORY
>>>   *	%CAIRO_STATUS_FILE_NOT_FOUND
>>>   *	%CAIRO_STATUS_READ_ERROR
>>> + *      %CAIRO_STATUS_PNG_ERROR
>>>   *
>>>   * Alternatively, you can allow errors to propagate through the drawing
>>>   * operations and check the status on the context upon completion
>>> @@ -799,6 +802,7 @@ cairo_image_surface_create_from_png (const char *filename)
>>>   *
>>>   *	%CAIRO_STATUS_NO_MEMORY
>>>   *	%CAIRO_STATUS_READ_ERROR
>>> + *      %CAIRO_STATUS_PNG_ERROR
>>>   *
>>>   * Alternatively, you can allow errors to propagate through the drawing
>>>   * operations and check the status on the context upon completion
>>
>> In both of the above hunks: The code around uses a tab while your patch adds spaces.
> 
> I also noticed I missed a case statement in cairo-spans.c. I will fix
> both issues before pushing.
> 
>> With the above fixed:
>>
>> Reviewed-by: Uli Schlachter <psychon at znc.in>
>>
>> (Should there also be matching unit test, similar to what my patch did?)
> 
> Your patch should work with the expected error code changed to
> CAIRO_STATUS_PNG_ERROR. I'll leave it for you to push your test.

Let's just skip the tests then.
/me hides

> I've attached three more patches:
> 
> - Update the errors in the utils/* files. They have not been updated for
> a while.
> 
> - Add CAIRO_STATUS_FREETYPE_ERROR for libfreetype errors. It now returns
> CAIRO_STATUS_NO_MEMORY for out of memory and CAIRO_STATUS_FREETYPE_ERROR
> for any other error.
> 
> - Add CAIRO_STATUS_WIN32_GDI_ERROR for GDI errors. I didn't try to
> isolate out of memory errors from other errors as it is not clear to me
> if a GDI out of memory error means "out of heap memory" or "out of some
> internal resource in Windows". If the latter it is more helpful to
> return a GDI error than to return NO_MEMORY.

All look good to me as well. Feel free to push with my R-b.

Cheers,
Uli
-- 
A normal person is just someone you don't know well enough yet.
 - Nettie Wiebe