The support for a custom message in the attribute 'deprecated' have been
added only from gcc 4.5.0 and not 4.0, so the check was updated accordingly
to avoid compilation error with version 4.0 to 4.4.
Reported-by: Josip Deanovic <djosip+news@linuxpages.net>
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The original code was using double precision floating point to perform the
color corrections for the creation of the standard colormap. This precision
is not necessary because color coding is 16 bits anyway, and on some
architecture the double precision comes with a cost.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
To preserve the accuracy of the operation, the C standard request that the
mathematical operation is performed using double precision, but in many
case this is not necessary so this patch fixes a few constants to avoid
that conversion.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As pointed by Coverity, in the "Configuration" panel of WPrefs there are
some images loaded, but if the images cannot be loaded correctly then the
returned NULL pointer can crash the application as it is dereferenced in
further function calls.
To solve this case, this patch is adding a NULL pointer check in the
functions RScaleImage (wrlib) and WMCreatePixmapFromRImage (WINGs), so both
can accept that NULL pointer to also return NULL, which means the existing
check for "icon == NULL" in the WPrefs code will be useful.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
This script should have been removed a long time ago, in favour of
pkg-config which has to be present anyway to compile Window Maker.
This patch makes the script print a warning about it, and now calls itself
pkg-config, instead of being generated; the man page was also updated
accordingly (and spell-checked).
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
Instead of having to maintain manually the map file for LD with the list of
symbols that we want to keep in the WRaster library (the purpose is to hide
internal symbols), this patch implements a script that parses the public
API header and extracts automatically the list of names that are supposed
to be visible to the user.
The goal is to reduce possible human errors, like for example the function
RCombineAlpha that was forgotten from the list, yet still keep the map file
feature which is considered a good practice (it reduces the risk for name
clash and the risk of mis-use of internal stuff from the library).
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
For code maintainability, it is better to have a single definition of the
constant than many ones dispatched in many places. In addition, we try to
count on the constant M_PI that can be defined by the header <math.h> if
possible because it may have the best accuracy for the platform.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
We have a few directories with source codes that we tell configure to
prepare, but we do not actually want them built during normal operations
(tests and examples only).
However, there are some special targets brought by automake which still
need to see them, so this patch adds these directories to the list, but
only for these rules, we keep them unvisited by the normal build process.
The wanted side effect of this is that now "make distcheck" works untill
the end as expectable.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As reported by Andrew, the compilation of the WRaster broke because
there was an API change in libgif v5.1 versus the v5.0 (something had been
forgotten for DGifCloseFile to be easily used in wrappers for dynamic
languages).
Now, if we have detected that we're in 5.x release, we use the GIFLIB_MINOR
macro to see what the function prototype is (this macro was introduced only
in 4.1.6 so we cannot fully rely on it to detect the version of the
library).
The possible error code is not used because at the place we use the
function we would not be able do do anything more meaningful with it.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
It seems that the argument to the attribute 'deprecated' is a novelty of
gcc 4.x, the attribute introduced in gcc 3.x did not take any parameter.
As Solaris 10 is providing gcc 3.4.3, this patch updates the detection in
the public API header to make it compile properly.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
cppcheck is reporting:
[wrlib/tests/testgrad.c:148]: (error) Memory leak: color_name
but seems some other variables were never freed.
This patch is cleaning them property.
WRaster:
- new function 'RShutdown'
- removed flag 'optimize_for_speed' from RContext
- new functions '...(operate_xxx)', '...(flip_image)'
WUtil:
- new function 'wutil_shutdown'
- new macro 'wlengthof'
WINGs:
- new function 'WMReleaseApplication'
- new function 'WMCreateScaledBlendedPixmapFromFile'
(And maybe a few more I missed)
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The commit #d1f9b801 removed it because the code behind it is actually
useless, unfortunately removing the flag itself from the structure is
breaking the compatibility of the API.
To avoid problems, this patch re-introduces the flag in the structure, but
not the code behind it so it is useless, and adds an attribute so that gcc
will report its use as deprecated, to help application migration.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
It is generally not a good idea to have an API with a high number of
functions because it adds complexity for user and for maintainability,
so both function have been "merged" into a single RFlipImage with
a parameter to specify what flip is expected.
As a bonus, the function can perform both flips at once if wanted.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As it does not cost anything in the current code to not add this channel,
it is then probably a good idea to keep the output image with the same
format as the input image, this avoid wasting +33% of memory for something
that may be at best unused and at worst will induce extra cost when
manipulating the image.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
Instead of calling all variables with variants of 'image', changed
the name to reflect what image it is about for clarity.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
The function RRotateImage contains code to handle efficiently the 4 simple
angles 0, 90, 180 and 270 degrees, which makes it a long function.
This patch separate the code for the different cases into dedicated
functions so the main function's code ends up being simple (aka: easier to
understand/review/maintain).
As a side effect, the function for the 180 degree function is not static
because it can be reused to flip an image both horizontally and vertically.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
When defining enums as types instead of simple enums allows to use these
types at the places where the corresponding enum values are expected, then
allowing the compiler to check that, potentially reporting incorrect use
of values to the user.
This patch adds the type for the gradient style for RRender*Gradient.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
When defining enums as types instead of simple enums allows to use these
types at the places where the corresponding enum values are expected, then
allowing the compiler to check that, potentially reporting incorrect use
of values to the user.
This patch adds the types for the drawing operations used by the functions
who drawing shapes into an RImage.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
When defining enums as types instead of simple enums allows to use these
types at the places where the corresponding enum values are expected, then
allowing the compiler to check that, potentially reporting incorrect use
of values to the user.
This patch adds types for the configuration fields in the RContextAttributes
structure.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
Using local function prototype redefinition is dangerous, now a single
prototype is seen in all the files that are concerned.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
To be consistent with RCreateXImage and because it should not be up to the
caller to handle this, when XGetImage returns a NULL pointer then the
function RGetXImage will also return NULL instead of an RXImage structure
with a NULL pointer.
This consistent behaviour helps fixing a memory leak in WMaker reported by
Coverity (#50125).
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As pointed by Coverity, this value is never used, the assignation has no
effect, so Coverity reports that it is unnecessary complexity that is not
compatible with code maintainability.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As pointed by Coverity, there's a free on the content of this buffer at
function's end, but if the image's width or height is lower than 1 then we
could reach this free before allocating anything, thus having an
uninitialised pointer.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As pointed by Coverity, environment variables cannot be considered
trustworthy ("tainted"), so we need to make sure the values provided are
within reasonable bounds for safe operations.
Seized the opportunity to try to provide clearer name for the constants
defining the default value.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As pointed by Coverity, the behaviour of fopen/fread/fclose in case of
error is not really what the macro RETRY assumes. So the macro is removed
and appropriate action is implemented.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As pointed by Coverity, the shift operation performed for color-to-pixel
transform may not behave as well as expected because of the types used
(Coverity pointing suspicious sign extension because of int type being
involved).
The new code tries to leave no open interpretation to the compiler.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
As pointed by Coverity, the shift operation performed for color-to-pixel
transform may not behave as well as expected because of the types used
(Coverity pointing suspicious sign extension because of int type being
involved).
The new code tries to leave no open interpretation to the compiler.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
With a set of images i was able to detect that
the flip functions was shifting the image by 1px.
The 90 and 270 degrees rotation were not working as
the functions were also mirroring the img.
The patch is also fixing the C style based on checkpatch.
when the rotation angle value passed to RRotateImage is a modulo of 90,
the function is working well but in other cases the rotateImage()
function is called. That last function is half implemented but
the half already implemented part is also segfaulting (use the testrot.c
to replay the crash).
So the patch is moving the 'not implemented' msg to disable the whole
rotateImage() function and comment it out.
Result of 'malloc' is converted to a pointer of type 'unsigned char',
which is incompatible with sizeof operand type 'char'
Signed-off-by: Amadeusz Sławiński <amade@asmblr.net>
The library uses internally a cache of tables to convert image to different
depths, there is now an internal function 'r_destroy_conversion_tables' to
free them.
Instead of calling 'MagickWandGenesis' every time an image is loaded, only
call it the first time, and similarly do not call 'MagickWandTerminus' if
the library has not been used before.
Signed-off-by: Christophe CURIS <christophe.curis@free.fr>
