You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

1298 lines
40 KiB

2009-01-14 Fridrich Strba <>
* gen_art_config.c: remove
* add
*, use a shell script together
with autotools to create the art-config.h. This way is cross-compile
friendly and gives the same result.
2008-01-30 Alexander Larsson <>
Release 2.3.20
2007-03-01 Frederic Crozat <>
reviewed by: Dom Lachowicz <>
* art_misc.h: Fix header when included in C++.
Patch from Goetz Waschk and Laurent Montel.
2007-02-28 Kjartan Maraas <> Release 2.3.19
2007-02-28 Kjartan Maraas <>
* Fix a typo introduced with the bug fix
for bug #131478. Noticed by Yanko Kaneti.
2007-02-26 Kjartan Maraas <>
* 2.3.18.
2006-12-06 Kjartan Maraas <>
* .cvsignore:
Port to automake 1.9. Patch from Christian Persch.
Closes bug #131478
2005-10-26 Alexander Larsson <>
Cygwin build fixes from Cygwin Ports maintainer (#317562)
2005-10-05 Dom Lachowicz <>
* art-misc.[ch]: Turn memory-managing macros into proper functions
instead of preprocessor wrappers around malloc. Win32 ships multiple
runtimes, and an art_alloc'd block from one runtime can't be
art_free'd by a different runtime.
* libart.def: Export the 3 new functions
2005-05-27 Alexander Larsson <>
* art_vpath_bpath.c: (art_vpath_render_bez):
Handle the case where start and endpoint are very near
in a better way. (#301750)
Patch from Bernhard Herzog
2005-03-24 Tor Lillqvist <>
* art_misc.h (ART_GNUC_PRINTF, ART_GNUC_NORETURN): Use __format__,
__printf__ and __noreturn__ to avoid warnings in case "format",
"printf" or "noreturn" are #defined to something else. For example
GNU libintl.h (at least some versions of it) #defines printf as
2005-01-24 Kjartan Maraas <>
* art_render_gradient.c: (art_render_gradient_radial_render):
* testart.c:
* testuta.c: (main): Fix some compiler warnings etc.
2005-01-21 Kjartan Maraas <>
* Release 2.3.17
2004-03-22 Michael Meeks <>
* (EXTRA_DIST): add libart.def.
2004-03-04 Glynn Foster <>
Add uninstalled pkg-config file.
Tue Sep 09 16:30:31 2003 George Lebl <>
* Release 2.3.16
* art_render_gradient.c: revert the double comparison test in the
asserts and comment out the two asserts causing crashes
everywhere, see bug #121850
2003-09-02 Alexander Larsson <>
* 2.3.15
2003-08-26 Morten Welinder <>
* art_render_gradient.c (art_render_gradient_linear_render_8): Fix
vertical gradients. Fix yet more floating point comparisons using
!=. For the record, Gnumeric is a heavy user.
2003-08-14 Tor Lillqvist <>
Some nitpicking changes for Win32:
* Declare the variables from the library as
dlimport/dllexport on Win32.
* Don't use -lm on Win32. Use the libart.def file.
* makefile.msc: Update to generate same name DLL and import
library as auto* and libtool. Add missing objects. (But dunno if
it is useful to maintain this file. It isn't distributed in the
source tarballs (should it be?). And, if it were distributed, and
the intention was to support random people really building libart
with MSVC, one should distribute also prebuilt config.h(.win32)
and libart-features.h(.win32) files.)
* (DEVZIP): Fix typo.
(DLLDIR): libtool 1.5 installs DLLs in $(prefix)/bin, so look
there, too.
* libart.def: Add missing entry points.
* testuta.c: To check that the export/import of the version
variables work, print out them in a comment. Undefine
LIBART_COMPILATION before including libart-features.h.
2003-08-08 Alexander Larsson <>
* 2.3.14
Fri Jul 25 12:29:35 2003 George Lebl <>
* art_render_gradient.c (art_render_gradient_linear_render_8)
(art_render_gradient_linear_render) (art_render_gradient_linear)
(art_render_gradient_radial_render) (art_render_gradient_radial):
Redo the checks where float was compared by == or != to using
the EPSILON define. Also copy the ArtGradientLinear and
ArtGradientRadial into the source structure, pretending that
these are constants that will never change or be freed by
the caller is utterly evil and in fact for librsvg it is
not constant. This fixes some more very random crashes
when using librsvg with libart (which seems to be the
only usage of the gradient stuff)
Fri Jul 18 12:57:36 2003 George Lebl <>
* art_render_gradient.c: Fix more comparison-of-doubles by == bugs,
this code is uber ugly. Should fix the fairly random crashes
on asserts I've been having.
2003-07-11 Michael Meeks <>
* Version 2.3.13
2003-07-11 Federico Mena Quintero <>
* art_svp_ops.c (art_svp_minus): impl.
Tue Jul 08 01:15:02 2003 George Lebl <>
* art_render_gradient.c: fix comment as pointed out by alex
Tue Jul 08 01:13:48 2003 George Lebl <>
* art_render_gradient.c (art_render_gradient_linear_render_8):
when we wish to find the current segment and we go beyond the
last stop due to float fun, use the last segment as that's
really what we want. Avoids a very abrupt assert death.
2003-05-05 Alexander Larsson <>
Bump to 2.3.12
2003-04-24 Alexander Larsson <>
* art_uta_vpath.c (art_uta_from_vpath):
Don't silently stomp on memory on bad vpaths.
2003-04-11 Alexander Larsson <>
* art_svp_vpath_stroke.c (render_seg):
Handle cases when dmr2 is very small better.
2003-04-10 Alexander Larsson <>
* art_svp_wind.c (x_order_2):
Handle horizontally aligned segments.
2002-11-25 Alexander Larsson <>
* Bump to 2.3.11
2002-10-16 John Harper <>
* art_render_svp.c (art_render_svp_callback,
art_render_svp_callback_span, art_render_svp_callback_opacity,
art_render_svp_callback_opacity_span): if no runs would
normally be emitted, but start is greater than zero, emit a
single run covering the entire width of the rendered region
2002-08-18 Havoc Pennington <>
* hardcode aclocal-1.4/automake-1.4 so that users with
both automake 1.6 and 1.4 installed get the right automake. Means
compilation from CVS will now require the latest automake 1.4
release, or manually creating symlinks called "automake-1.4" and
2002-07-01 Alexander Larsson <>
* Release 2.3.10
2002-07-01 Alexander Larsson <>
Bump to 2.3.10
* art_svp_intersect.c (art_svp_intersect_add_seg):
Initialize seg->wind_left to zero, this avoids
uninitialized memory read later in art_svp_intersect_horiz_commit.
2002-06-24 Alexander Larsson <>
Release 2.3.9.
2002-06-24 Alexander Larsson <>
* Bump to 2.3.9
2002-06-04 Balamurali Viswanathan <>
* Added -lm to libart_lgpl_2_la_LIBADD
fixes bug 75711 (Jacob's suggestion)
2002-05-08 Raph Levien <>
* art_uta_vpath.c (art_uta_add_line): Fixes very subtle
edge case found by Federico Mena Quintero: (96, 96) -
(96.220200017562348, 93.034868598919431). Previously, through
numerical error, xn was a hair to the left, throwing off
the Bresenham iteration.
2002-03-08 Tor Lillqvist <>
* Minor changes for build on Win32. Call
AC_LIBTOOL_WIN32_DLL. Set automake conditionals OS_WIN32 and
* art_affine.c: Include art_misc.h for M_PI, which not necessarily
is in math.h.
* On Win32, build and install import libraries, both
for gcc and for MSVC (if available).
* New file. Used to build runtime and developer
package for Win32.
2002-02-06 Laszlo Peter <>
* art_render_gradient.c: copy alloca hacks from glib/galloca.h
2002-02-02 Alexander Larsson <>
* art_render.c:
Add special case for art_render_composite_8 for two cases
that are very common when rendering SVGs with librsvg.
I'm sure these can be optimized further, but i'm feeling
very slow today.
2002-02-01 Alex Larsson <>
* art_render_gradient.c:
Add optimized case for depth==8, n_channels == 3
* test_gradient.c:
Test case for gradients. Uses gtk+, so not built by
2002-01-10 Anders Carlsson <>
* Release 2.3.8
2002-01-04 Anders Carlsson <>
* Bump version to 2.3.8.
2002-01-03 Darin Adler <>
* Add art_render_mask.[ch].
* art_bpath.h:
* art_rect_svp.h:
* art_rect_uta.h:
* art_render.h:
* art_render_gradient.h:
* art_render_mask.h:
* art_render_svp.h:
* art_rgb_svp.h:
* art_svp_intersect.h:
* art_svp_ops.h:
* art_svp_point.h:
* art_svp_render_aa.h:
* art_svp_vpath.h:
* art_svp_vpath_stroke.h:
* art_svp_wind.h:
* art_uta.h:
* art_uta_rect.h:
* art_uta_ops.h:
* art_uta_svp.h:
* art_uta_vpath.h:
* art_vpath_bpath.h:
* art_vpath_dash.h:
* art_vpath_svp.h:
Fix includes so that each header includes what it needs.
* art_affine.c:
* art_alphagamma.c:
* art_bpath.c:
* art_gray_svp.c:
* art_misc.c:
* art_pixbuf.c:
* art_rect.c:
* art_rect_svp.c:
* art_rect_uta.c:
* art_render.c:
* art_render_gradient.c:
* art_render_mask.c:
* art_render_svp.c:
* art_rgb.c:
* art_rgb_a_affine.c:
* art_rgb_affine.c:
* art_rgb_affine_private.c:
* art_rgb_bitmap_affine.c:
* art_rgb_pixbuf_affine.c:
* art_rgb_rgba_affine.c:
* art_rgb_svp.c:
* art_rgba.c:
* art_svp.c:
* art_svp_intersect.c:
* art_svp_ops.c:
* art_svp_point.c:
* art_svp_render_aa.c:
* art_svp_vpath.c:
* art_svp_vpath_stroke.c:
* art_svp_wind.c:
* art_uta.c:
* art_uta_ops.c:
* art_uta_rect.c:
* art_uta_svp.c:
* art_uta_vpath.c:
* art_vpath.c:
* art_vpath_bpath.c:
* art_vpath_dash.c:
* art_vpath_svp.c:
Fix order of includes so that the corresponding header is
included first, to test that each header includes what it
2002-01-02 Darin Adler <>
* art_alphagamma.h:
* art_gray_svp.h:
* art_pixbuf.h:
* art_render.h:
* art_rgb.h:
* art_rgba.h:
Add some missing includes.
2001-12-26 Raph Levien <>
* art_svp_intersect.c: More fixes to intersector. When tops
coincide, full ordering test is performed (with breaking),
rather than just comparing b from line eq. Also, had a test
for x >= backwards in art_svp_intersect_add_point. Thanks
to Bruce Q. Hammond for test cases.
2001-11-21 Raph Levien <>
* art_svp_intersect.c (art_svp_intersect_break): Changed break
logic systematically so that new breaks are always in order,
rather than allowing them to become out of order, then try to fix
them up later. It's no doubt possible to come up with cases in
which this reduces precision, but it simplifies life nicely, so I
did it.
* (art_svp_intersect_horiz): Initialize (a, b, c) values of hs to
avoid UMR's when they are tested later.
* art_svp_ops.c (print_ps_vpath): Changed the coordinate
transform so that it's done in the PostScript rather than
the C code that outputs the coordinates. This makes it easier
to reconstruct the vector path from the debug output.
2001-11-16 Alex Larsson <>
* art_rect.c (art_drect_svp_union, art_drect_svp):
Don't call art_drect_union() in these functions, since
it considers zero-width or zero-height svg segments to
be empty. This causes it to think i.e. rectangular svps
are empty.
2001-11-07 Raph Levien <>
* art_svp_intersect.c (art_svp_intersect_horiz): Fixed important
logic bug, failing to test crossings of neighbors when inserting
horizontal lines. Also changed printf to art_dprint.
* art_misc.c (art_dprint): Added function for debug printing,
so verbose intersector output doesn't have to go through printf.
* art_misc.h (ART_USE_NEW_INTERSECTOR): I've turned this on
now, as the new intersector certainly seems to be performing
better than the old one now.
2001-10-31 Anders Carlsson <>
* Release 2.3.7
2001-10-15 Raph Levien <>
* art_svp_intersect.c (art_svp_intersect_horiz): Minor
logic fix so that horiz segments successfully cross
zero length segments in the active list.
(art_svp_intersect_test_cross): Flags indicating whether to
do add_point (potentially breaking neighbors) to left and
to right.
(art_svp_intersect_insert_cross): Provide ART_BREAK_LEFT and
ART_BREAK_RIGHT flags to art_svp_intersect_test_cross,
depending on direction of search.
(art_svp_intersect_advance_cursor): Provide flags (allow
both left and right breaking) to test_cross.
2001-10-15 Raph Levien <>
* art_svp_intersect.c (CHEAP_SANITYCHECK): Added an inexpensive
sanitycheck to detect multiple insertions of a segment into the
horiz list.
(art_svp_writer_rewind_add_point): Avoid breaking lines below
their bottom point.
(art_svp_intersect_test_cross): Handle cases correctly where
intersection point matches y0 of left or right segment. These _do_
happen in real world examples. Also, do add_point on newly
inserted intersection point.
2001-10-14 Raph Levien <>
* art_svp_intersect.c (art_svp_intersect_add_point): Fixed
rather subtle logic bug that misplaced insertion point
when seg argument was NULL.
2001-10-11 Raph Levien <>
* art_svp_render_aa.c (art_svp_render_aa_iter_step): Got rid
of qsort of steps, and now keep the step list in sorted order.
Further, we combine duplicate steps with the same x value,
which bounds the size of the step list to x1 - x0, so we
don't need to dynamically resize it. Thanks greatly to
Bruce Q. Hammond for the original version of this patch.
2001-10-09 Raph Levien <>
* art_svp_intersect.c (art_svp_intersect_test_cross): Breaks
bottom part of line segments in "too close" cases.
2001-10-09 Raph Levien <>
* art_svp_intersect.c (art_svp_writer_rewind_add_point): Fixed
bbox computation.
(art_svp_intersector): Handle degenerate case where input
SVP has 0 segments.
* art_svp_intersect.h: Moved definition of art_svp_intersector
inside #ifdef __cplusplus, so it links properly against C++
2001-10-09 Raph Levien <>
* art_svp_intersect.c (art_svp_intersect_break): Handle
case when break y equals sweep line correctly. Also adds
first try at winding number sanitychecker, but that makes
too many false positives.
2001-10-07 Raph Levien <>
* art_svp.c (EPSILON): Set to zero if new intersector is
in use - we want svp's to be in strict sorted order.
* art_svp_intersect.c (art_svp_intersect_test_cross): Explicitly
check that top points are equal, and swap immediately if b is out
of order.
(art_svp_intersect_horiz): Break segments that intersect
horizontal lines. Now passes "two squares with offset" test.
2001-10-05 Raph Levien <>
* art_svp_intersect.c: Major changes to accommodate
horizontal lines. Intersections of horizontal lines
aren't fully processed, but should work a lot better
than before.
* testart.c: Minor tweaks. testpat now frees memory
so it can be run under memprof to detect leaks.
2001-10-03 Raph Levien <>
* art_svp_intersect.c (art_svp_intersect_advance_cursor):
Made test_cross for inserted segments common between
intersection processing and cursor advance, and also took
care of a case that hadn't been handled before. Also added
invariant sanitychecker for debugging purposes.
2001-10-02 Raph Levien <>
* art_svp_ops.c: ART_USE_NEW_INTERSECTOR variants of svp
ops changed to do shallow free of merged svp.
2001-10-01 Raph Levien <>
* art_svp_intersect.c:
* art_svp_intersect.h:
* art_misc.h:
* art_svp_wind.h: First commit of new intersector code. It is
turned off by default, but can be enabled by #defining
* art_svp_ops.c: Make svp ops use new intersector if enabled.
* art_svp_vpath_stroke.c: Make vpath stroking use new intersector
if enabled.
* testart.c: New test case for intersector.
Wed Sep 26 03:48:13 2001 George Lebl <>
* Release 2.3.6
Wed Sep 26 03:11:40 2001 George Lebl <>
* gen_art_config.c: Fix 64bit issues, cast sizeof return when
using %d to print it.
2001-09-13 Havoc Pennington <>
* rename library to libart_lgpl_2
* (Cflags): move includes
* move includes
* delete, rename libart-config
(EXTRA_DIST): don't install m4 files
(libart_lgplincdir): move headers
2001-08-03 Michael Meeks <>
* Version 2.3.5 for the API freeze.
2001-07-12 Darin Adler <>
* art_affine.c: (art_affine_expansion): Fix handling of
negative numbers. We ran into this bug a while back when
figuring out why librsvg couldn't handle certain svg files.
2001-07-12 Darin Adler <>
* art_misc.h: Change art_expand macro so it's a single
statement, using the do while (0) trick, which gets rid
of some warnings.
* art_pixbuf.c: Add a missing include.
* art_vpath_svp.c: (art_vpath_from_svp): Initialize a
variable to avoid a compiler warning.
* gen_art_config.c: Add a missing include.
2001-03-24 Martin Baulig <>
Applied the patch from Alexander Larsson which was sitting
in gnome-libs/patches/libart.diff since February.
[FIXME: Alex, can you please provide a ChangeLog?])
* art_rgb_a_affine.[ch]: New files.
2001-01-07 Hans Breuer <Hans@Breuer.Org>
* art_misc.c : embryonic change to use libart_lgpl on win32
* libart.def : new file, exported functions for win32 dll
* makefile.msc : handwritten for MSVC compiler
2000-09-30 Martin Baulig <>
* Provide pkg-config script.
* Create libart-2.0.pc from
* (pkgconfig_DATA): Install the libart-2.0.pc
script in `$(libdir)/pkgconfig'.
2000-08-15 Raph Levien <>
* art_render.c (art_render_image_solid_negotiate): Only
sets ART_IMAGE_SOURCE_CAN_COMPOSITE when a compositing
callback is selected. Previously was causing segfaults on
non-alpha images. Thanks to Leonard Rosenthol for spotting
the bug.
Fri Jun 30 22:56:58 2000 Raph Levien <>
* art_render.c (art_render_composite): Fixed a bug that caused
it to ignore the alpha setting. Also art_render_composite_8().
2000-06-01 John Sullivan <>
* art_svp_render_aa.c: (art_svp_render_aa_iter_step):
Made it build by correcting struct member name from
Raph's previous checkin.
Wed May 31 11:10:58 2000 Raph Levien <>
* art_svp_render_aa.c (art_svp_render_aa_iter_step): Updated
n_steps_max in iter structure after steps reallocation.
Tue May 30 10:33:13 2000 Raph Levien <>
* art_svp_render_aa.c (art_svp_render_aa_iter_step): Fixed not
updating iter->steps when steps gets reallocated.
2000-05-30 Pavel Cisler <>
* art_rgba.c:
Make it build -- fix a broken include.
Tue May 30 00:09:21 2000 Raph Levien <>
* art_render_gradient.c (art_render_gradient_setpix): Fixed
an off-by-one loop error.
Mon May 29 15:00:39 2000 Raph Levien <>
* Moved relevant .h files into HEADERS stanza.
Mon May 29 13:48:49 2000 Raph Levien <>
This is a fairly major commit, as it adds both the new, modular
rendering architecture and gradients. Quite a bit of the code
feels like "reference code" now, in that it is (hopefully)
correct, but not necessarily very fast. In addition, there remain
a few things not done, including the use of SVP's as non-driver
mask sources. AlphaGamma and filter level also remain
unimplemented. No compositing modes other than ART_NORMAL are
implemented. All in good time!
* added -Wmissing-prototypes flag. Bumped version
number to 2.3.0.
* art_render.h:
* art_render.c: Added new rendering architecture.
* art_render_svp.h:
* art_render_svp.c: Added renderers to use SVP's as mask
sources in new rendering architecture.
* art_render_gradient.h:
* art_render_gradient.c: Added linear and radial gradients
as image sources in new rendering architecture.
* art_rgba.h:
* art_rgba.c: Added functions for manipulating and compositing
RGBA pixels.
* art_svp_wind.c: Added static to trap_epsilon(), #ifdef'd out
* art_uta_ops.c: Added #include "art_uta_ops.h".
* art_uta_rect.c: Added #include "art_uta_rect.h".
* art_uta_svp.h: fixed __ART_UTA_SVP_H__ name.
* art_misc.h: Added ART_GNUC_NORETURN attribute, added that
to the prototype for art_die(). Added "static" to function
declarations to avoid warnings when compiled with
* testart.c: Added gradient test.
Thu May 25 23:30:39 2000 Raph Levien <>
* art_svp_render_aa.h:
* art_svp_render_aa.c: Added art_svp_render_aa_iter functions,
suitable for iterative rendering of an svp, one scan line at a
* Bumped version to 2.2.0.
Tue May 16 15:03:35 2000 Raph Levien <>
* art_rgb_pixbuf_affine.c: Included corresponding .h file.
* art_rgb_pixbuf_affine.h: Put recursive #includes inside
* art_gray_svp.c:
* art_rgb_svp.c: Explicit casts for callback data. Also removed
"render the steps into tmpbuf" comment.
* gen_art_config.c:
* Added code to automatically generate an
art_config.h file, to be installed in libart's include dir. This
file defines ART_SIZEOF_{CHAR,SHORT,INT,LONG} and art_u{8,16,32}.
* art_misc.h: Moved definition of art_u8 and art_u32 into
art_config.h. Added GCC printf format attributes.
* art_svp_wind.c (traverse): Fixed UMR bug here. The function
isn't actually used, so it's just for cleanliness.
2000-04-18 Lauris Kaplinski <>
* art_affine.c (art_affine_to_string): Replaced snprinf with
art_ftoa to avoid localisation of generated numbers
2000-04-18 ERDI Gergo <>
* art_rgb_pixbuf_affine.h: Included the ArtPixBuf declaration
Fri Apr 14 16:33:55 2000 Raph Levien <>
* art_svp_wind.c (art_svp_uncross, art_svp_rewind_uncrossed):
Fixed uninitialized memory reads when inserting new segment into
* art_bpath.c (art_bpath_affine_transform): Made it avoid
potential uninitialized memory reads when not all the coordinates
are needed. Thanks to Morten Welinder for spotting both of these
2000-04-05 Raph Levien <>
* art_svp_wind.c: Make "colinear" warnings go to stderr instead
of stdout. Of course, when I finish the new intersector, these
will go away entirely.
2000-04-04 Raph Levien <>
* art_uta_vpath.c (art_uta_add_line): Fixed bug that was causing
segfaults on alphas. Thanks to msw for localizing it.
2000-01-17 Raph Levien <>
* art_svp_vpath_stroke.c (art_svp_vpath_stroke): Typo in api
header (thanks rak).
2000-01-16 Timur Bakeyev <>
* Instead of jumping between srdir and builddir just process
all the auto* staff in srcdir. In fact, just saying the same things in
other words.
2000-01-10 Elliot Lee <>
*, *.h: Add rather bad hacks to the header files to allow compilation
* Distribute
2000-01-09 Raph Levien <>
art_rgb_pixbuf_affine.c, art_rgb_rgba_affine.c, art_rgb_svp.c,
art_svp.c, art_svp_ops.c, art_svp_point.c, art_svp_render_aa.c,
art_svp_vpath.c, art_svp_vpath_stroke.c, art_svp_wind.c,
art_uta.c, art_uta_ops.c, art_uta_rect.c, art_uta_svp.c,
art_uta_vpath.c, art_vpath.c, art_vpath_bpath.c, art_vpath_dash.c,
art_vpath_svp.c: Added API documentation.
Fri Sep 24 17:53:21 1999 Raph Levien <>
* art_svp_render_aa.c (art_svp_render_insert_active): Avoid
reading undefined memory (thanks to Morten Welinder).
1999-09-19 Raph Levien <>
* art_pixbuf.c (art_pixbuf_duplicate): Added a duplicate function
at the request of Michael Meeks.
1999-09-11 Raph Levien <>
* art_affine.c (art_affine_to_string): Tightened the predicate for
matching rotate-only affines, which was too weak. Thanks to lewing
for spotting it!
1999-09-01 Raph Levien <>
* art_affine.c, art_alphagamma.c, art_bpath.c, art_gray_svp.c,
art_misc.c, art_pixbuf.c, art_rect.c, art_rect_svp.c,
art_rect_uta.c, art_rgb.c, art_rgb_affine.c,
art_rgb_bitmap_affine.c: Updates to api doc headers.
1999-08-24 Raph Levien <>
* art_affine.c, art_alphagamma.c, art_alphagamma.h, art_bpath.c,
art_bpath.h, art_gray_svp.c, art_misc.c, art_pixbuf.c,
art_pixbuf.h, art_point.h, art_rect.c, art_rect.h: Added api
documentation headers.
* testart.c: Added "dash" test, for testing the vpath_dash
* art_rgb_pixbuf_affine.h: Fixed the #ifdef for conditional
inclusion. Thanks to Kristian Hogsberg Kristensen for spotting
the bug.
1999-08-24 Raph Levien <>
* art_svp_render_aa.c (art_svp_render_aa): Added some tests to
avoid NaN for infinite slopes, which were causing problems on
Alphas. Closes bug #1966.
1999-08-20 Federico Mena Quintero <>
* Fixed library's libtool version number.
1999-08-03 Larry Ewing <>
* art_vpath_dash.c (art_vpath_dash): fix a bug/typo that was causing
certain paths to loop infinitely.
1999-07-28 Raph Levien <>
* art_vpath_dash.[ch]: Added a function to add a dash style
to vpaths. It is tested, but has a couple of rough edges (see
code for details).
* testart.c: added tests for the new vpath_dash functionality.
* added art_vpath_dash.[ch] files.
1999-07-26 Raph Levien <>
* art_rgb.c (art_rgb_fill_run): fixed incorrect test for
big-endianness. Thanks to Michael Zucchi for spotting it.
Fri Jul 16 23:42:59 1999 Tim Janik <>
* art_affine.c (art_affine_flip): flip translation matrixes as well, by
inverting matrix[4] if (horz) and inverting matrix[5] if (vert).
Fri Jul 16 23:03:26 1999 Tim Janik <>
* art_pixbuf.[hc]: deprecated art_pixbuf_free_shallow(), people should
always free pixbufs with art_pixbuf_free() and use the _dnotify variants
for specific destruction behaviour.
added art_pixbuf_new_rgb_dnotify() and art_pixbuf_new_rgba_dnotify()
which allow for a destruction notification function. (this involved
adding two extra pointers to the ArtPixBuf structure, and removal of
the recently introduced int flags field).
Mon Jul 12 01:13:23 1999 Tim Janik <>
* art_affine.[hc]: added art_affine_equal(), which checks two
matrixes for equality within grid alignment.
Fri Jul 9 17:50:19 1999 Tim Janik <>
* art_affine.[hc]: added art_affine_flip() to flip a matrix horizontally
and/or vertically, or just copy it.
added art_affine_shear() to setup a shearing matrix.
Tue Jul 6 19:03:39 1999 Tim Janik <>
* art_pixbuf.h: added an int flags; member to the end of the
structure, it currently only holds information about whether the
pixels member should be freed. (raph: i think flags is more generic
than free_pixels, so we can reuse that field if further demands popup
in the future).
* art_pixbuf.c:
(art_pixbuf_new_const_rgb): new functions that prevent the pixels
member from being freed upon art_pixbuf_free ().
(art_pixbuf_free): only free the pixels member if it is non-NULL and
1999-07-02 Raph Levien <>
* art_vpath_bpath.c (art_vpath_render_bez): Bad bad uninitialized
* added compile warnings. Guess why :)
1999-06-28 Raph Levien <>
* art_svp_point.h:
* art_svp_point.c: Added methods for insideness and distance
testing, very useful for ::point methods in canvas items.
* testart.c: test code to exercise the art_svp_point functions.
* Additional entries for art_svp_point.
1999-06-28 Raph Levien <>
* art_svp_render_aa.c (art_svp_render_aa): Subtle boundary
case in realloc code -- was causing nasty segfaults.
Wed Jun 23 15:05:43 1999 Raph Levien <>
* art_rgb_svp.c (art_rgb_svp_alpha_opaque_callback): Missed a
case in the anti-segfault crusade. Thanks lewing!
Wed Jun 23 11:16:42 1999 Raph Levien <>
* art_rgb_svp.c: Made these routines so they won't segfault even
if alpha is out of range. Of course, that begs the question of
fixing the render routines so they won't _make_ alpha go out of
range, but all in good time.
Fri Jun 18 17:32:34 1999 Raph Levien <>
* art_vpath_bpath.c (art_bez_path_to_vec): Switched to a new
adaptive subdivision algorithm, which (finally!) takes flatness
into account. This should result in both smoother curves and
faster operation.
Sun Jun 13 21:07:20 1999 Raph Levien <>
* art_svp_wind.c (art_svp_rewind_uncrossed): Made the winding
rule logic even more correct :). I somehow missed the fact that
a clockwise path should be given a winding number of zero;
last night's commit tried to make it -1 (which worked for the
test cases I was using).
Sun Jun 13 01:23:14 1999 Raph Levien <>
* art_svp_wind.c (art_svp_rewind_uncrossed): Change to winding
rule logic so that it correctly handles the case where the
leftmost segment is negative.
* (libart_lgplinc_HEADERS): made art_svp_wind.h
a public headerfile. This is needed for the bpath canvas item.
I'm not sure this is the correct way to do it, but it will do
for now.
* art_vpath_bpath.h:
* art_vpath_bpath.c (art_bez_path_to_vec): Added const to arg.
* art_vpath_bpath.h: Embarrassing typo.
* art_bpath.h: Minor tweaks to the #include paths. It is now
consistent with the other header files.
Wed Jun 9 20:24:45 1999 Raph Levien <>
* art_svp_vpath_stroke.c: Added all remaining line join and cap
types, including round, which takes flatness into account. Several
new internal functions (art_svp_vpath_stroke_arc) and added
flatness argument to a few internal functions. I might want to
change the BEVEL join type to MITER for very small turn angles
(i.e. within a flatness threshold) for efficiency.
* art_misc.h: Added M_SQRT2 constant.
Wed Jun 2 21:56:30 1999 Raph Levien <>
* art_svp_vpath_stroke.c (art_svp_vpath_stroke_raw): Made the
closed path detection capable of PostScript semantics (i.e. it
now senses the difference between ART_MOVETO and ART_MOVETO_OPEN).
* art_svp_vpath_stroke.c (art_svp_vpath_stroke_raw): it now
filters out successive points that are (nearly) coincident. This
fixes some of the crashes and hangs, including Tim Janik's
singularity (trying to stroke MOVETO 50, 50; LINETO 50, 50; END).
* art_svp_wind.c (art_svp_rewind_uncrossed): added a test to
correctly handle empty input svp's.
* art_svp_wind.c (art_svp_uncross): added a test to correctly
handle empty input svp's.
Sun Jan 17 20:53:40 1999 Jeff Garzik <>
* art_affine.c:
Include string.h for memcpy.
* art_svp_vpath.c:
Remove conflicting static func definition.
* art_uta_svp.c:
Include art_vpath_svp.h for func definition.
Mon Jan 4 12:47:47 1999 Raph Levien <>
* art_bpath.c (art_bpath_affine_transform): Stupid misnaming
of this function (forgot the "art_").
Thu Dec 31 09:04:23 1998 Raph Levien <>
* art_affine.c (art_affine_rectilinear): Added this function.
* art_rect.c (art_drect_affine_transform): Corrected the name (it
was right in the .h). Also made it work with non-rectilinear
transforms, while I was at it.
Thu Dec 17 11:58:24 1998 Raph Levien <>
* art_alphagamma.h:
* art_alphagamma.c: The real code for alphagamma.
Wed Dec 16 14:18:46 1998 Raph Levien <>
* art_alphagamma.h:
* art_alphagamma.c: Added. At present, it only contains a dummy
stub. When the real code is added, it supports doing alpha
compositing in a gamma-corrected color space (suppressing
* art_pixbuf.h:
* art_pixbuf.c: Added. This is a virtualization layer over
a few different kinds of image formats.
* art_rgb_pixbuf_affine.h:
* art_rgb_pixbuf_affine.c: Added. Supports compositing of
generic images over an rgb buffer.
* art_affine.h:
* art_affine.c (art_affine_expansion): Added this function,
which reports the exact scale factor in the case of rotation,
scaling, and transformation (an approximate number in the
case of shearing or anamorphic distortion).
* art_misc.h:
* art_misc.c (art_warn): Added.
* art_rgb_affine.h:
* art_rgb_affine.c: Added alphagamma argument (not yet implemented).
* art_rgb_affine_private.c: Fixed typo bug that was causing
repaint problems for nonsquare images.
* art_rgb_bitmap_affine.h:
* art_rgb_bitmap_affine.c: Major speed improvement, probably fixed
correctness while I was at it. Added alphagamma argument (not yet
* art_rgb_svp.h:
* art_rgb_svp.c: Added alphagamma argument (only implemented
in aa case, not yet alpha case).
* art_vpath.c: increased perturbation to 2e-3, because the old
value (1e-6) was too small.
* testart.c: added alphagamma.
* added new files
Sun Dec 27 21:45:03 1998 Raph Levien <>
* art_rect.h:
* art_rect.c: Added DRect versions of the basic ops (double
rather than int).
* art_rect_svp.h:
* art_rect_svp.c: Added. This computes the bounding box of
an svp.
Wed Dec 16 14:18:46 1998 Raph Levien <>
* art_alphagamma.h:
* art_alphagamma.c: Added. At present, it only contains a dummy
stub. When the real code is added, it supports doing alpha
compositing in a gamma-corrected color space (suppressing
* art_pixbuf.h:
* art_pixbuf.c: Added. This is a virtualization layer over
a few different kinds of image formats.
* art_rgb_pixbuf_affine.h:
* art_rgb_pixbuf_affine.c: Added. Supports compositing of
generic images over an rgb buffer.
* art_affine.h:
* art_affine.c (art_affine_expansion): Added this function,
which reports the exact scale factor in the case of rotation,
scaling, and transformation (an approximate number in the
case of shearing or anamorphic distortion).
* art_misc.h:
* art_misc.c (art_warn): Added.
* art_rgb_affine.h:
* art_rgb_affine.c: Added alphagamma argument (not yet implemented).
* art_rgb_affine_private.c: Fixed typo bug that was causing
repaint problems for nonsquare images.
* art_rgb_bitmap_affine.h:
* art_rgb_bitmap_affine.c: Major speed improvement, probably fixed
correctness while I was at it. Added alphagamma argument (not yet
* art_rgb_svp.h:
* art_rgb_svp.c: Added alphagamma argument (only implemented
in aa case, not yet alpha case).
* art_vpath.c: increased perturbation to 2e-3, because the old
value (1e-6) was too small.
* testart.c: added alphagamma.
* added new files
Mon Dec 14 00:16:53 1998 Raph Levien <>
* art_affine.c (art_affine_to_string): re-added the "scale" method
that was accidentally deleted before check-in.
* added new files
Sun Dec 13 00:52:39 1998 Raph Levien <>
* art_affine.h:
* art_affine.c: Added. Everything you ever wanted to do with an
affine transform. Especially check the functions that generate
concise PostScript strings for affine transforms.
* art_filterlevel.h: A simple enum for selecting filtering
* art_rgb_affine.h:
* art_rgb_affine.c (art_rgb_affine): Added. This function
composites an (opaque) rgb image over an rgb pixel buffer. At
present, it's slow and only nearest neighbor filtering is enabled.
* art_rgb_rgba_affine.h:
* art_rgb_rgba_affine.c: Analogous, but for compositing rgba
* art_rgb_bitmap_affine.h:
* art_rgb_bitmap_affine.c: Analogous, but for compositing bitmap
* art_rgb_affine_private.c (art_rgb_affine_run): Added. This is
a common function used by all the rgb_affine modules to move
testing for source image bbox out of the inner loop.
* added the new files
* testart.c: exercise the image compositors
Wed Dec 9 23:36:35 1998 Raph Levien <>
* art_vpath.c (art_vpath_perturb): Made it deal correctly
with closed paths (the MOVETO and closing LINETO have to
* art_svp_wind.c: Made the bbox calculations for the resulting
svp's correct.
* art_svp.h:
* art_svp.c: The art_svp_seg_compare function moved here, because
it's required in art_svp_ops.
* art_svp.c (art_svp_add_segment): It now does bbox calculations.
* art_svp_ops.h:
* art_svp_ops.c: Added. Populated with basic union, intersection,
and diff functions.
* art_vpath_svp.h:
* art_vpath_svp.c: Added. Populated with a function to convert
from sorted to unsorted vector paths
* added the new files
* testart.c: exercise the stroke outline and vector path
1998-12-08 Herbert Valerio Riedel <>
* art_svp_wind.c: added #include <string.h> for memcpy()
Sun Dec 6 22:15:12 1998 Raph Levien <>
* art_svp_wind.[ch], art_svp_vpath_stroke.[ch]: Added, but it
doesn't work yet. These will do stroke outline and basic
vector ops like union, intersect, etc.
* art_svp_render_aa.c: Added a simple speedup based on bbox
culling. I will want to do more speedups, but none of this is
necessary for the freeze.
* art_svp_vpath.c: Fixed some bugs in the art_svp_from_vpath in
cases where there is more than one subpath.
* art_vpath.h:
* art_vpath.c (art_vpath_perturb): Added this function. This will
help cheat as long as the basic vector ops have numerical
stability problems.
Fri Dec 4 18:00:38 1998 Raph Levien <>
* art_svp_render_aa.c (art_svp_render_aa): Changed the api
slightly, to guarantee that the steps are all within the range
from x0 (inclusive) to x1 (exclusive).
* art_gray_svp.c, art_gray_svp.h: Added. Populated with functions
to render into a simple graymap.
* art_rgb.c, art_rgb.c: Added. Populated with fill_run and
run_alpha methods.
* art_rgb_svp.c, art_rgb_svp.h: Added. Populated with functions to
render into an RGB buffer, and to composite over an RGB buffer.
* added art_gray_svp, art_rgb, and art_rgb_svp.
* testart.c: test the color and layered rendering.
Mon Nov 30 01:30:25 1998 Raph Levien <>
* testart.c: added vector path rendering stuff. Some of it needs
to go out of the test framework and into the module, but the
api hasn't settled down entirely yet (in the real code, all
x's in the step field are within bounds).
* art_svp_render_aa.c, art_svp_render_aa.c.h: added.
* art_svp_vpath.c, art_svp_vpath.h: added.
* art_pathcode.h: added ART_MOVETO_OPEN (libart uses an
ART_MOVETO_OPEN code at the beginning to indicate an open path,
while PostScript uses the lack of a closepath at the end).
* art_vpath_bpath.c, art_vpath_bpath.h: fixed it up, added
flatness arg to api.
* added new source files.
Wed Nov 25 17:19:44 1998 Raph Levien <>
* art_svp.h, art_svp.c: added, basic constructors for sorted
vector paths.
Sun Nov 22 23:21:09 1998 Raph Levien <>
* (libart_lgplincdir): Fixed stupid bug in naming of
the variable.
Sun Nov 22 21:41:13 1998 Raph Levien <>
* art_uta_vpath.c: moved art_uta_union into art_uta_ops.
* art_uta_ops.[ch]: added, populated with art_uta_union.
Thu Nov 19 00:19:40 1998 Raph Levien <>
* added
* added creation of, added -version-info
* added LIBART_VERSION_INFO, support for
* libart.m4: updated version history :)
Wed Nov 18 18:15:20 1998 Raph Levien <>
* (LIBART_VERSION): set this, so that libart-config
--version now works.
Wed Nov 18 16:50:58 1998 Raph Levien <>
* libart.m4: added (just copied from esound)
*, added support for libart-config
* added (mostly copied from esound)
Tue Nov 10 12:43:30 1998 Raph Levien <>
* Getting the library in shape for initial checkin to CVS.