I demand that Bob Friesenhahn may or may not have written...

> On Sun, 8 Aug 2010, Christopher Martin wrote:
>> I suspect that the crash is related to this problem:
>> https://www-old.cae.wisc.edu/pipermail/octave-maintainers/2010-January/014891.html
>> Thus it may be a xine-lib bug after all, as I don't see any
>> InitializeMagick() - or DestroyMagick() for that matter - calls in
>> xine-lib's code. I would guess that libxinevdec/image.c would be the
>> appropriate place to add them.

The right places are probably these:
  InitializeMagick() immediately prior to NewMagickWand();
  DestroyMagick() in dispose_class().

This assumes that it's safe to call DestroyMagick() without having called
InitializeMagick().

If this works, I'll patch and re-upload, reassigning and closing this bug
report along the way. (Sven, or anybody else who is able to trigger the
problem, could you test?)

> The next release of GraphicsMagick will make
> InitializeMagick()/DestroyMagick() totally thread safe.  In general it is
> safe to invoke InitializeMagick() just prior to use (and any number of
> times) so that life is easier within a library context.  The overhead of
> invoking InitializeMagick() is quite low since if the library is already
> initialized, the function immediately returns.

http://www.imagemagick.org/download/www/api/deprecate.html#InitializeMagick
says that InitializeMagick() is deprecated, which is... interesting, in this
context. (Being able to use either is somewhat important.)

> I am considering using GCC's shared library contructor/destructor semantics
> but am not sure how reliable that might be.

Well... as far as xine-lib is concerned, if that works with dlopen() &
dlclose()...

-- 
| Darren Salt            | linux at youmustbejoking | nr. Ashington, | Toon
| using Debian GNU/Linux | or ds    ,demon,co,uk    | Northumberland | back!
| + Output less CO2 => avoid massive flooding.    TIME IS RUNNING OUT *FAST*.

Scotty: "Damage control is easy. Reading Klingon, that's hard."



-- 
To UNSUBSCRIBE, email to [email protected]
with a subject of "unsubscribe". Trouble? Contact [email protected]

Reply via email to