[GM-core] Fwd: Magick++ Bug Report

[Hyrum W. <hw...@go...>]

See below for a bug in ImageMagick that Dirk thinks may also be in
GraphicsMagick.

---------- Forwarded message ----------
From: Dirk Lemstra <di...@le...>
Date: Tue, Feb 17, 2015 at 6:28 PM
Subject: Re: Magick++ Bug Report
To: Hyrum Wright <hw...@go...>


 Hyrum,

You might also want to contact the maintainer of GraphicsMagick. Their code
has the same issue.

Dirk

On 18-02-2015 00:07, Lexie Parsimoniae wrote:

From: Hyrum Wright <hw...@go...>
Reply-To: Hyrum Wright <hw...@go...>
X-Mailer: PHP/5.3.3
Origin: 72.14.228.1

On the 6.9.0 branch, Magick++/lib/Blob.cpp and Magick++/lib/Image.cpp don't
do proper locking.  Specifically, the line of the form:

  Lock(&_imgRef->_mutexLock);

or

  Lock(&_blobRef->_mutexLock);

are wrong.  These lines look like they are attempting to use RAII to ensure
the mutex is unlocked when the relevant functions return, but instead these
lines create a *temporary* unnamed lock that is immediately released.  The
proper fix looks something like:

  Lock lock(&_imgRef->_mutexLock);

I count 6 instances of this bug in Blob.cpp and 8 in Image.cpp.  These were
found using the misc-unused-raii check from clang-tidy.

It doesn't look like this problem exists on trunk, since the locking has
been moved to BlobRef.cpp and ImageRef.cpp and made manual.

-Hyrum