Re: error from Magick being caught as warning

[Carnë Draug]

On 17 December 2013 19:36, Alexander Hansen <address@hidden> wrote:
> On 12/17/13 11:31 AM, Carnė Draug wrote:
>>
>> Hi
>>
>> I noticed some time ago that ErrorCoder exceptions from Magick++ [1]
>> are being caught as warnings [2] but I don't understand why. The error
>> means that the image can't be read so imvec is empty leading to a
>> segfault when trying to access its first element (does not happen in
>> imread which checks the number of images read, but does in imfinfo).
>> There is no attempt to an alternative image reading method (there's in
>> private/__imread__.m a try/catch but that's waiting for an error
>> anyway).
>>
>> Why is this happening? Can I change that to actually give an error?
>>
>> I'm also CC'ing Sųren (previous maintainer of the image package),
>>
>> Teemu Ikonen (person who made the commit [3]), and Andy Adler (which
>> I'm guessing to be the original author since he's the only copyright
>> owner of the file since 2002 [4] even though the first commit dates
>> 2005).
>>
>> Carnė
>>
>>
>> [1] http://www.graphicsmagick.org/Magick++/Exception.html
>> [2]
>> http://hg.savannah.gnu.org/hgweb/octave/file/a86d608c413c/libinterp/dldfcn/__magick_read__.cc#l655
>> [3] https://sourceforge.net/p/octave/image/ci/d756a7b6d533
>> [4]
>> http://hg.savannah.gnu.org/hgweb/octave/file/a86d608c413c/libinterp/dldfcn/__magick_read__.cc#l4
>>
>
> Is this related to http://savannah.gnu.org/bugs/?38280 as well?

I don't think so. You mention the problem goes away if you don't get
the "View" and "TileName" info. The segfault here happens way before,
right after readImages, when accessing the first element of imvec.

By the way, the 3.8 version won't be reading those fields (also won't
be reading LongFormat, TotalColors, Matte, ModulusDepth, and
QuantizeColors. See details in cset ba79ba4e83ab). Could you please
try to build the RC1, and confirm that your problem went away?

Carnë