On Thu, Nov 7, 2013 at 1:29 PM, another gol another...@gmail.com wrote:
It could be that all of my monitors have a 1.4 gamma, but really, I never
saw it looking wrong on any monitor, there must be a reason Windows
defaults to 1.4. I strongly suspect that the 2.2 standard gamma is either a
Antti, you're sooo off topic
Please make an effort to come up with a subject line next time and start a
new thread.
Alexei
On Wed, Nov 13, 2013 at 1:52 AM, Antti Lankila alank...@bel.fi wrote:
Dave Arnold darn...@adobe.com kirjoitti 11.11.2013 kello 23.35:
Nothing stupid that I can see :-)
On Fri, Mar 14, 2014 at 11:18 AM, Sean McBride s...@rogue-research.comwrote:
Attached are 3 more patches, this time to fix compiler warnings (not
static analyzer warnings). None of them are serious. The biggest fixes
most -Wmissing-variable-declarations warnings, where some functions that
On Tue, Mar 18, 2014 at 2:37 AM, Werner LEMBERG w...@gnu.org wrote:
The patch 0004 is false positive. The variable 'hit' is set to zero
when entering the loop so 'p_last' and 'p_first' are assigned almost
immediately.
Thanks for the analysis. However, initializing doesn't hurt, and
On Tue, Mar 18, 2014 at 2:37 AM, Werner LEMBERG w...@gnu.org wrote:
The patch 0004 is false positive. The variable 'hit' is set to zero
when entering the loop so 'p_last' and 'p_first' are assigned almost
immediately.
Thanks for the analysis. However, initializing doesn't hurt, and
Perhaps, this is http://savannah.nongnu.org/bugs/?38662
___
Freetype-devel mailing list
Freetype-devel@nongnu.org
https://lists.nongnu.org/mailman/listinfo/freetype-devel
Hi Rodger,
Your tests show that Freetype stroker relies on
On Sat, May 24, 2014 at 12:49 AM, Rodger Combs rodger.co...@gmail.comwrote:
While profiling libass in OSX's Instruments.app, I've found that
Freetype's stroker tends to come up as a very time-consuming routine, and
that most of the
On Wed, Mar 19, 2014 at 2:36 AM, Werner LEMBERG w...@gnu.org wrote:
Arg... I am not sure if the patch is even correct. Aren't we
resetting the variable in each loop cycle? This wasn't the case in
the original code where the values reused, was it?
Oops! You are right. Thanks for
On Tue, Jun 17, 2014 at 10:29 AM, Sean McBride s...@rogue-research.com
wrote:
On Mon, 16 Jun 2014 20:46:44 -0400, Alexei Podtelezhnikov said:
Arg... I am not sure if the patch is even correct. Aren't we
resetting the variable in each loop cycle? This wasn't the case
Indeed, the loop body is exited and reentered in each iteration so the
scope is lost each time. It is only thanks to the reuse of stack that the
code worked. The last fix by Werner is correct. I would initialize to -1
though for consistency or not initialize at all.
On Thu, Jun 19, 2014 at 12:40
Hi all,
The following will roughly double the space of optimal calculations.
Comments?
Alexei
diff --git a/src/base/ftcalc.c b/src/base/ftcalc.c
index 6e65583..baeb810 100644
--- a/src/base/ftcalc.c
+++ b/src/base/ftcalc.c
@@ -363,15 +363,15 @@
/*
On Fri, Jul 4, 2014 at 12:54 AM, Werner LEMBERG w...@gnu.org wrote:
The following will roughly double the space of optimal calculations.
Comments?
Well, at the same time it reduces the available range for `c' by more
than 50%.
I was greedy with the space for a and b. With a + b 2*46340
On Fri, Jul 4, 2014 at 6:12 AM, Werner LEMBERG w...@gnu.org wrote:
Well, at the same time it reduces the available range for `c' by
more than 50%.
I was greedy with the space for a and b. With a + b 2*46340
instead, c can use the same old limit. We can expand c a lot further
by
On Fri, Jul 4, 2014 at 3:30 PM, Behdad Esfahbod beh...@behdad.org wrote:
On 14-07-04 03:09 PM, Alexei Podtelezhnikov wrote:
You force me to think, man. To avoid any overflow it is sufficient to
satisfy
this inequality
a + b 2 * sqrt(X - c/2)
where X is 2^31 - 1. Using Taylor
a + b 92681 - (c 16) + ((c+ 235935) 18)
How about the simpler:
a + b = 92681 - (c 16)
I like this!
Pushed.
___
Freetype-devel mailing list
Freetype-devel@nongnu.org
https://lists.nongnu.org/mailman/listinfo/freetype-devel
On Sat, Jul 5, 2014 at 2:05 AM, Werner LEMBERG w...@gnu.org wrote:
a + b 92681 - (c 16) + ((c+ 235935) 18)
How about the simpler:
a + b = 92681 - (c 16)
I like this!
Pushed.
Thanks a lot! However, please expand the comment, since the involved
mathematics is not trivial. For
On Mon, Jul 7, 2014 at 4:24 PM, Behdad Esfahbod beh...@behdad.org wrote:
On 14-07-07 03:01 PM, Alexei Podtelezhnikov wrote:
Many architectures have special instructions for the most significant
bit: BSR on x86, CLZ on ARM. Some compilers have builtin functions for
that too: __builtin_clz
On Fri, Aug 8, 2014 at 4:56 PM, Behdad Esfahbod beh...@behdad.org wrote:
* src/base/ftcalc.c (FT_MSB): Utilize gcc builtins.
introduced undefined behavior, since __builtin_clz and __builtin_ctz have
undefined behavior if passed in value is zero.
This can actually happen with degenerate
* src/pfr/pfrobjs.c, do you think count can be 0 here?
FT_UIntpower = 1 FT_MSB( count );
I doubt very highly that undefined behavior can mean a number between
0 and 31. So we'll get power = 0 if count = 0, which is fine too.
Every case is safe, or so it seems to me.
On Fri, Aug 8, 2014 at 4:56 PM, Behdad Esfahbod beh...@behdad.org wrote:
commit 177982e933ed6f2ab96163e14f4267f8abe89efd
Seriously though, my commit does not change much, both FT_MSB and
ft_highpow2 used to return nonsense on zero input even before this
commit. I think I am going to apply some
On Mon, Aug 11, 2014 at 11:55 AM, Behdad Esfahbod beh...@behdad.org wrote:
On 14-08-11 10:15 AM, Alexei Podtelezhnikov wrote: On Fri, Aug 8, 2014 at
4:56 PM, Behdad Esfahbod beh...@behdad.org wrote:
commit 177982e933ed6f2ab96163e14f4267f8abe89efd
Seriously though, my commit does not change
Ok. I still do not think that we should fix the macro, because there
is no good value to return. Returning zero for zero is bad too. One
should not call MSB or the builtins with zero argument, just like one
should not divide by zero. So I'll try to apply fix the callers.
It's sometimes more
Werner,
FT_MulFix optimized like this on line 543:
if ( ua = 2048 ub = 1048576L )
ua = ( ua * ub + 0x8000UL ) 16;
Of course, when we anchor 2048 and use unsigned multiplication, it
should safely have been
if ( ua = 2048 ub = 2097135L )
but that's not my point. I do not know
Under FT_LONG64 FT_DivFix truncates the result at 32 bits, see
ftcalc.c:271. Interestingly, FT_MulFix and FT_MulDiv do not do that,
although it also is possible for them to exceed 32 bits in the result.
Which behavior would be correct? IMHO a function specified as FT_Long
should not truncate at
On Sat, Aug 16, 2014 at 2:08 PM, Behdad Esfahbod beh...@behdad.org wrote:
Scratch this part. the existing limit is more than enough for all practical
purposes. It's pixel-size divided by upem in 16.16. So you need a font size
1 pixels before you surpass a 1048576 scale.
Graham, Behdad.
On Sat, Aug 16, 2014 at 2:06 PM, Behdad Esfahbod beh...@behdad.org wrote:
I think main usage pattern for this function is to scale values from font
space to Fixed device space. Font-space values are typically 2048 because
of TrueType commonly uses upem=2048 and Type1 uses 1000, so 2048
Under linux on x86_64, the long type is 8 bits. Why do we disable FT_LONG64?
/* A 64-bit data type will create compilation problems if you compile*/
/* in strict ANSI mode. To avoid them, we disable its use if __STDC__ */
/* is defined. You can however ignore this rule by defining
On Wed, Aug 20, 2014 at 10:20 AM, Werner LEMBERG w...@gnu.org wrote:
Under linux on x86_64, the long type is 8 bits. Why do we disable
FT_LONG64?
/* A 64-bit data type will create compilation problems if you compile*/
/* in strict ANSI mode. To avoid them, we disable its use if
I want to move CURRENT_Ppem outside of the point loops. It seems to
work fine. Anything wrong with that?
diff --git a/src/truetype/ttinterp.c b/src/truetype/ttinterp.c
index 9491533..082114a 100644
--- a/src/truetype/ttinterp.c
+++ b/src/truetype/ttinterp.c
@@ -7490,7 +7490,7 @@
FT_ULong
On Sun, Aug 24, 2014 at 10:24 PM, Alexei Podtelezhnikov
apodt...@gmail.com wrote:
I want to move CURRENT_Ppem outside of the point loops. It seems to
work fine. Anything wrong with that?
This is OK, since the ppem (measured along the projection vector)
doesn't change within a single call
On Wed, Aug 27, 2014 at 12:42 AM, Werner LEMBERG w...@gnu.org wrote:
Furthermore, it turns out that MPS instruction is *very* *very*
common. Why do we call CURRENT_Ppem here? The comment suggest that
the decision was arbitrary and, perhaps, wasteful. I see a huge
number of trivial
In memory of Paul Alexi
I have just committed significant optimization of ft_div64by32, one of
the more expensive calculations that we sometimes do on 32-bit
platforms. The trick was to do as much as we can with 32-bit division
before embarking on
Some time ago James Cloos proposed...
I tested a wide array of gcc cross compilers as well as clang/llvm.
In each case the machine code this version produces is as good as
the hand-tuned assembly and uses the trick that adding 0x7FFF when
the product is negative causes the same
On Fri, Oct 3, 2014 at 3:27 PM, Alexei Podtelezhnikov
apodt...@gmail.com wrote:
... this patch:
http://cgit.freedesktop.org/~cloos/freetype2/commit/?id=94dc0c20554f3a3d709711f2c43ac65ab7c68afe
Never mind. Right-shifting negative numbers is implementation defined
(ANSI or C11), so
Werner,
For a while I was staring at
ttinterp.c: B = B * 64 / ( 1L CUR.GS.delta_shift );
delta_shift is FT_Short, so it can be negative, which is undefined
behavior. Shouldn't we change the type to FT_UShort?
The second question is how often do you see delta_shift larger than 6?
I am
On Mon, Oct 13, 2014 at 11:27 AM, Alexei Podtelezhnikov
apodt...@gmail.com wrote:
The second question is how often do you see delta_shift larger than 6?
I am basically thinking about this optimization:
B = B * ( 1L ( 6 - CUR.GS.delta_shift ) );
I think I found the answer:
The legal
On Tue, Oct 14, 2014 at 1:09 AM, Werner LEMBERG w...@gnu.org wrote:
The legal range for delta shift is zero through six. Negative
values are illegal.
Yep, since the smallest unit is 1/(2^6) = 1/64th px.
What should freetype do with larger values? Should we set delta_shift,
but do nothing in
On Mon, Oct 13, 2014 at 7:04 PM, Behdad Esfahbod beh...@behdad.org wrote:
The first two commits remove about 1600 lines combined.
The file contained bytecode description in plain English and consumed
stack specs. I will miss it.
+1 if you preserve English descriptions and stack specs.
On Mon, Oct 13, 2014 at 7:04 PM, Behdad Esfahbod beh...@behdad.org wrote:
I was checking it out today and noticed that there are two main loops: default
big switch, and non-default function-dispatch array.
Both are likely implemented through jump tables with similar
performance. Whether the
Hi all,
According to
https://developer.apple.com/fonts/TrueType-Reference-Manual/RM05/Chap5.html#S45ROUND
The grid period tables of the 45-degree S45ROUND is actually sqr(2)
*larger* than that of SROUND. You might get different impression from
reading more cryptic Microsoft files. So shouldn't
Werner,
Admittedly, I enjoy reading those Apple docs too much. So I found
other problems:
Here
https://developer.apple.com/fonts/TrueType-Reference-Manual/RM02/Chap2.html#rounding_operations
the phase is added back *before* checking if rounding changed the
sign. MS is in agreement here. Freetype
So I found other problems:
In addition Round_To_Double_Grid does half-grid instead.
val = ~31
should be
val = ~127
I'll start patching tonight.
___
Freetype-devel mailing list
Freetype-devel@nongnu.org
In addition Round_To_Double_Grid does half-grid instead.
val = ~31
should be
val = ~127
Brrr, sorry, I'm wrong, freetype is right here.
___
Freetype-devel mailing list
Freetype-devel@nongnu.org
#define DO_S45ROUND \
-SET_SuperRound( 0x2D41, args[0] ); \
+SET_SuperRound( 0x5A82, args[0] ); \
CUR.GS.round_state = TT_Round_Super_45; \
CUR.func_round = (TT_Round_Func)Round_Super_45;
Good question. I
On Wed, Oct 22, 2014 at 6:06 AM, John Emmas john...@tiscali.co.uk wrote:
Hi guys, I updated freetype from git master (this morning) and I'm getting
these errors when compiling cff.c (technically, when compiling cffobjs.c):-
Freetype did not used to put such tweaking knobs in the options file.
Anyways, since VC 2010 is a frequently used compiler, we have to find
a workaround. Behdad, any ideas?
I do not have the compiler but I would try braces around {the
constants}. Then you can use it to initialize a temp array to transfer
in the appropriate place at last.
FT_Int params[] =
gcc warning:
In file included from /home/apodtele/freetype2/src/cff/cff.c:27:0:
/home/apodtele/freetype2/src/cff/cffobjs.c: In function 'cff_driver_init':
/home/apodtele/freetype2/src/cff/cffobjs.c:1091:22: warning: typedef
'static_assert_darkening_parameters' locally defined but not used
Am I right that advance is always rounded to pixels no matter if the
glyph is hinted or not? I wonder if this should be reflected in
lsb_delta and rsb_delta, perhaps just the latter. What do you think?
Alexei
___
Freetype-devel mailing list
Werner,
What will it take to change FT_Outline_Funcs API? I am talking about
shift and delta. Freetype does not use them and sets them zero. Google
search outside seems to show the same. This API is so inflexible that
it is useless.
On the other hand, subpixel rendering scales outlines 3 times
On Tue, Nov 18, 2014 at 9:05 PM, Behdad Esfahbod beh...@behdad.org wrote:
It was reported to me by my colleague Doug Felt (CC'ed) that FT_MulFix takes
FT_Long parameters as defined in freetype.h, but several inline
implementations of it in ftcalc.h take FT_Int32 arguments.
I cannot imagine
On Wed, Nov 19, 2014 at 12:32 AM, Werner LEMBERG w...@gnu.org wrote:
I was also wondering if FT_Long should be defined synonymous to
FT_Int32, etc.
typedef FT_Long FT_Int32
Hmm. FT_Int32 is *exactly* 32bit, while FT_Long is simply a typedef
to `long' and thus *at least* 32bit, depending
These PaxHeaders.xxx directories were created when using the standard
BSD tar program. When I downloaded and used GNU tar the build worked
fine. Can freetype be packaged using a more portable tar method?
Quoting from http://packages.ubuntu.com/de/lucid/bsdtar
GNU tar support. Libarchive
Will have a look whether I can follow your patch, doing the
reduction by myself.
Humm. Did you get to take a look at this by any chance?
It's on my TODO list
If all relevant compilers implement the switch as a jump table, indeed
there is no need to have an internal surrogate
Over the past couple of days I developed a tool for testing FreeType in
multi-threaded environments:
Your patches are now in FreeType's git repository. Thanks a lot!
IMHO, this is a big deal that needs to be released as 2.6.0.
___
Freetype-devel
We don't release unstable versions.
Thanks Alexey - so will there (eventually) be a branch in git which
corresponds to Release v2.5.x? Or does the branch called STABLE
automatically reflect whatever happens to be
Historically the most recent is the best. We don't have resources to
Why is it supposed to work? I do not follow.
#define INT_MIN 0x8000
This might be an unsigned integer constant according to standards.
When used, it may result in implicit type conversion to unsigned for
the rest of calculations and comparisons: the hell breaks loose. This
is a reason
On Fri, Feb 13, 2015 at 1:45 PM, Hin-Tak Leung
ht...@users.sourceforge.net wrote:
This might catch some interesting and hidden bugs.
BTW, mingw header defines it in include/limits.h as:
#define INT_MIN (-2147483647 - 1)
Freetype does not redefine *standard* INT_MIN. If someone decided to
On Thu, Jun 4, 2015 at 4:39 PM, Behdad Esfahbod beh...@behdad.org wrote:
On 15-06-04 12:19 PM, Behdad Esfahbod wrote:
On 15-06-04 12:15 PM, Alexei Podtelezhnikov wrote:
On Wed, Jun 3, 2015 at 4:28 PM, Behdad Esfahbod beh...@behdad.org wrote:
TrueType fonts have a 'non-zero winding' fill rule
On Wed, Jun 3, 2015 at 4:28 PM, Behdad Esfahbod beh...@behdad.org wrote:
TrueType fonts have a 'non-zero winding' fill rule, whereas Postscript has
'even-odd'. That's what causes the artifacts.
FT_OUTLINE_EVEN_ODD_FILL is never set by freeetype. So actually
FreeType always uses non-zero
Hi All,
It looks like FreeType supports BDF files with MS style bits_per_pixel
but not FontForge extension. See this:
http://fontforge.github.io/BDFgrey.html
Should we add it?
Alexei
___
Freetype-devel mailing list
Freetype-devel@nongnu.org
Hi All,
I think we abuse the suffix L in integer constants. I understand the
desire to make sure that very common 0x1L is 32-bit to represent a
unit in 16.16 fixed-point representation. On modern 64-bit systems
that actually becomes an unnecessary 64-bit constant. Perhaps the
suffix was used
On Wed, Jun 24, 2015 at 1:12 PM, Tom Bishop, Wenlin Institute
wen...@wenlin.com wrote:
I would have guessed that with a 64-bit compiler the resulting executable
code would be the same with or without L. (The definitions of types like
FT_Fixed and FT_Long would make more difference.)
There is this page:
http://fontforge.github.io/assets/old/Composites/index.html
Nice, thanks! After seeing this page now I can faintly remember that
I've seen it a looong time ago also :-)
I think George Williams wrote it. However his empirical equations on
this page and in the FreeType
if ( slot-format == FT_GLYPH_FORMAT_OUTLINE )
FT_Outline_EmboldenXY(..., ...-darken_x, 0);
This function still needs a bit of work around the corners, both
literally and figuratively speaking. Emboldening is not affine
transformation. It can introduce artifacts. The function has some
crude
No, unless FT_Outline_Embolden() does that, then yes ;)
`FT_Outline_Embolden' *does* change the advance width! You have to
save the metrics before the call, then restoring the original values.
Don't you mean experimental FT_Glyph_Embolden?
___
1. The function to compute how much darkening to apply is a crude copy-pasta
of the real thing in cff/cf2font.c
Is this the one with four parameters? I do not understand why it has to be so
complex. I am not a big fan unless someone can explain it?
Alexei A Podtelezhnikov, PhD
On Aug 24, 2015, at 6:11 AM, Behdad Esfahbod behdad.esfah...@gmail.com
wrote:
On 15-08-23 11:29 PM, Werner LEMBERG wrote:
`FT_Outline_Embolden' *does* change the advance width! You have to
save the metrics before the call, then restoring the original
1. The function to compute how much darkening to apply is a crude
copy-pasta of the real thing in cff/cf2font.c
Is this the one with four parameters? I do not understand why it has
to be so complex. I am not a big fan unless someone can explain it?
I haven't yet looked closely at the new
FT_Outline_EmboldenXY only changes the outline and its bounding box. The
dimensions of the letter is changed very predictably by xstrength and
ystrength. The outline is shifted by half of the strengths so that the
bottom-left of the bounding box stays in place. Makes sense?
Yes, this I
This effect:
https://www.flickr.com/photos/behdad/34769511/
still happens in my limited testing.
This looks like a text layout problem, which was traditionally outside
of FreeType's responsibilities. The extra ink is a result of bitmap
addition, which would be swept under the rug by strong
On Tue, Jun 30, 2015 at 3:47 AM, Werner LEMBERG w...@gnu.org wrote:
Uh, oh, this code is very fragile and not easy to understand... I
tried to remove some compiler warnings (from `clang
-Wsign-conversion', see below for the original compiler messages and
my bad attempt to fix them), and even
Hi All,
I have just committed a fast vector normalization procedure based on
Newton's iterations without square roots and divisions. It speeds up
emboldening by 40% and TrueType loading by 15% in my benchmarks with
times and arial at 16ppem.
There are a few places in the code that are
On Tue, Jul 28, 2015 at 1:26 AM, Werner LEMBERG w...@gnu.org wrote:
I still get one warning:
ftcalc.c:870:17: warning: implicit conversion changes signedness:
'int' to 'FT_UInt32' (aka 'unsigned int') [-Wsign-conversion]
l = 0x1 + (FT_Int32)( u * x + v * y ) / 0x1;
~
On Tue, Jul 7, 2015 at 11:20 AM, Werner LEMBERG w...@gnu.org wrote:
It is well known that offsets in composite or compound glyphs were
treated differently by AAPL and MSFT. The former scales them the
latter does not [...]
Do you have fonts to test?
No, unfortunately.
What about Apple GX
Hi All,
Type 1 fonts use FontMatrix to define scaling from character space to
user space. FreeType actually deduces units_per_EM from the yy
component of the matrix. This is the primary purpose of FontMatrix. Of
course, it can also be used to specify narrowing, expending, slanting,
rotation, or
On Sun, Aug 2, 2015 at 6:03 AM, Christian Demmer ch...@gcjd.org wrote:
On Sun, 02 Aug 2015 09:50:21 +0200 (CEST), Werner LEMBERG wrote:
I will also send you what I have.
I just searched my old stuff and I found Apple's QuickDraw GX v1.1.3
from 1995 containing a folder with Testing Fonts:
I can imagine that introducing some auxiliary variables help clean up
*and clarify* the issue. Compilers today are smart enough to optimize
them away, causing no loss in speed.
My last commit should remove all conversion warnings.
___
On Thu, Oct 22, 2015 at 10:34 PM, Alexei Podtelezhnikov
<apodt...@gmail.com> wrote:
> On Thu, Oct 22, 2015 at 7:44 PM, Nikolaus Waxweiler <madig...@gmail.com>
> wrote:
>>> My monitor is sRGB. Yours is likely too. We are looking at the corrected
>>> glyphs.
&
Dave Arnold wrote:
> I completely agree with Nikolaus. To state it a different way, correct
> anti-aliasing requires that the compositing (alpha blending) is done in a
> linear color space.
So I quoted your image
http://lists.nongnu.org/archive/html/freetype-devel/2013-10/msg00069.html
The bottom
On Thu, Oct 22, 2015 at 3:55 PM, Nikolaus Waxweiler wrote:
>> There is a cool image attached to this bug report:
>> https://bugreports.qt.io/browse/QTBUG-41590
>>
>> Gamma correction makes black text lighter and white text heavier,
>> i.e., has opposite effects, i.e., you
On Thu, Oct 22, 2015 at 12:00 PM, Nikolaus Waxweiler wrote:
> And gamma correction leads to better results with light text on dark
> background as far as I can see: https://bel.fi/alankila/lcd/
There is a cool image attached to this bug report:
On Thu, Oct 22, 2015 at 4:43 PM, Dave Arnold <darn...@adobe.com> wrote:
>
>
> On 10/22/2015 12:21 PM, Alexei Podtelezhnikov wrote:
>>
>> On Thu, Oct 22, 2015 at 12:00 PM, Nikolaus Waxweiler <madig...@gmail.com>
>> wrote:
>>>
>>> And gamma c
On Sat, Oct 17, 2015 at 4:46 PM, Nikolaus Waxweiler wrote:
>
> https://github.com/madig/freetype2/tree/stem-darkening-autohinter
>
Great work! A couple of suggestions though:
- I think stem darkening should be tested without gamma applied for
many reasons. Your monitor might
On Mon, Oct 26, 2015 at 10:26 AM, Alexei Podtelezhnikov
<apodt...@gmail.com> wrote:
> What would be best default value though? David Turner argued
> that it is gamma = 1.4
> (http://lists.nongnu.org/archive/html/freetype-devel/2004-06/msg00022.html).
> David Arnold also thinks
On Mon, Oct 26, 2015 at 10:26 AM, Alexei Podtelezhnikov
<apodt...@gmail.com> wrote:
> On Sun, Oct 25, 2015 at 2:35 PM, Nikolaus Waxweiler <madig...@gmail.com>
> wrote:
>> The world needs more linear alpha blending and gamma correction. Spread the
>> gospel!
>
&
On Sun, Oct 25, 2015 at 2:35 PM, Nikolaus Waxweiler wrote:
> The world needs more linear alpha blending and gamma correction. Spread the
> gospel!
We can start with changing the default gamma value in ftsting
(defaults to 2.0) and ftview (defaults to 1.0). They have always
http://lists.nongnu.org/archive/html/freetype-devel/2013-10/pngJZYanEzf8F.png
That image by Dave Arnold is an outstanding demonstration why gamma
correction is necessary on sRGB monitors. I could not help but show
this on a different scale, see attached. My recent favorite test is
this:
ftview
On Thu, Oct 22, 2015 at 7:44 PM, Nikolaus Waxweiler wrote:
>> My monitor is sRGB. Yours is likely too. We are looking at the corrected
>> glyphs.
>
>
> No we aren't, precisely because of our sRGB monitors. The bitmaps FreeType
> produces wouldn't need to be corrected if our
On Tue, Nov 10, 2015 at 7:43 AM, Jan Alexander Steffens
wrote:
> On Tue, Nov 10, 2015 at 1:32 PM, Markus Trippelsdorf
> wrote:
>> The big problem is that enabling linear blending is a compile time
>> option for Qt 5. There is no way to control it
On Mon, Nov 9, 2015 at 4:12 PM, Nikolaus Waxweiler wrote:
> So Alexei noted that he found stem darkening too strong. And it got me
> thinking again. As long as cairo/Qt5/Skia and others don't ship with LAB+GC...
Yes, I am afraid stem darkening is premature until gamma
> Any progress, Alexei? :)
Next release of freetype-demo tools will have starting gamma value of 1.8.
http://git.savannah.gnu.org/cgit/freetype/freetype2-demos.git/commit/?id=b48a54c1499debc39e73c1b024064ecb6d24d29e
___
Freetype-devel mailing list
On Wed, Nov 4, 2015 at 6:24 AM, Nikolaus Waxweiler wrote:
> Any progress, Alexei? :)
I am sitting here in total disbelief and utterly embarrassed by the
lack of gamma correction setting for text rendering in any major Linux
desktop environment. How could this happen? Pango
On Wed, Oct 7, 2015 at 4:33 PM, Behdad Esfahbod
wrote:
> Can you please make more quick-and-small releases for important fixes?
It does not matter considering that
Fedora 22 runs 2.5.5, which is progressive in comparison to
Android 5.1.1 runs 2.5.3
Ubuntu Vivid runs
On Thu, Oct 8, 2015 at 2:39 AM, Jan Engelhardt wrote:
>>Why distributions are more conservative about FreeType than kernel is beyond
>>me.
>
> If you have too old a kernel, some of your hardware may not
> activate/behave properly. If you have too old a freetype, the worst that
>
On Wed, Oct 7, 2015 at 2:57 PM, Kostya Serebryany wrote:
>> Yep. If you like, I can give you write access so that you can quickly
>> adjust the fuzzer function if necessary.
>
> Do you have a pre-commit code review process for this kind of changes?
>
We are just a responsible
On Tue, Oct 6, 2015 at 11:43 PM, Kostya Serebryany <k...@google.com> wrote:
> On Tue, Oct 6, 2015 at 7:59 PM, Alexei Podtelezhnikov <apodt...@gmail.com>
>> Anyhow, there is not much left to do or fuzz after
>> rendering, so do not render
>
> If you
On Tue, Oct 6, 2015 at 9:42 PM, Kostya Serebryany wrote:
> 51.76% repro_orig repro_orig [.] gray_render_scanline
That's been FreeType workhorse since 2.0. You gotta draw those glyphs
on a bitmap. It is funny you should mention how expensive divisions
there,
It is well known that offsets in composite or compound glyphs were
treated differently by AAPL and MSFT. The former scales them the
latter does not [...]
Do you have fonts to test?
No, unfortunately.
You do not have classic Geneva, Chicago, Monaco, and New York, do you?
We've got to find
Hi Toshiya and Werner,
It is well known that offsets in composite or compound glyphs were
treated differently by AAPL and MSFT. The former scales them the
latter does not
This page describes offset scaling
On Thu, Aug 27, 2015 at 9:02 AM, Werner LEMBERG w...@gnu.org wrote:
All of the sudden these 4 pair have an
incredible honor to be configurable. Why? What is so special about
them?
Apparently, nobody can even explain what they mean.
??? They are well documented!
The look of the function
101 - 200 of 932 matches
Mail list logo