[cairo] Fwd: Cairo 1.14 error on OS X 10.10.3 Yosemite

suzuki toshiya mpsuzuki at hiroshima-u.ac.jp
Fri May 29 13:00:52 PDT 2015


Dear Sirs,

Because 1.14.2 release has shown that the current regression
test for Quartz backend has a room for some improvement,
I propose to include a dynamic switch for legacy CGFont API
and new CTFont API in Quartz backend that was drafted by Andrea.

I guess, executing same binary on different version of Mac
OS X would be easier than the comparison by different cairo
version for same application. In fact, the change between
cairo-1.14.0 and cairo-1.14.2 (and 1.14.4) is not only by
Quartz backend. If an user finds an unexpected result in
newer release, there is a possibility that the difference is
not caused by the change of Quartz backend.

When we could establish better regression test, it would be
the time to reconsider dropping the legacy backend.

Any comment?

Regards,
mpsuzuki

Bryce Harrington wrote:
> On Wed, May 13, 2015 at 01:19:13AM +0000, Daven Quinn wrote:
>> Hi and thanks for your help with this issue. I built with HEAD and verified
>> that the results are correct.
>>
>> Since the issue has been rectified in HEAD, then I think it is resolved
>> from my perspective. If you think the solution is brittle, it may be better
>> to prioritize the older functions for now (I don't know how this change
>> affects compatibility with older versions of OS X). Either way works, as
>> long as the buggy behavior is not reverted to.
>>
>> What is the timeline for a release of a stable 1.14.3?
> 
> The next release is 1.14.4 btw.  Odd release numbers are reserved for
> devel snapshots.
> 
> As far as timing goes, I am going to focus on reviewing and pushing
> cairo patches for a few weeks, then push a release after.  So I'm
> figuring 1.14.4 may be in the July-ish timeframe.
> 
> Meanwhile, if there are any further Mac related changes you guys are in
> consensus that you want included, make sure to get patches in to me
> shortly.
> 
> Bryce
> 
>> Thanks,
>> Daven
>>
>>
>>
>> On Sun, May 10, 2015 at 9:38 AM suzuki toshiya <mpsuzuki at hiroshima-u.ac.jp>
>> wrote:
>>
>>> Hi,
>>>
>>> I could reproduce the issue, and I guess it could be
>>> fixed by upgrading the latest git (or future release
>>> of 1.14.3). My results are:
>>> http://gyvern.ipc.hiroshima-u.ac.jp/~mpsuzuki/cooling-scenarios_1.14.1.pdf
>>> (good)
>>> http://gyvern.ipc.hiroshima-u.ac.jp/~mpsuzuki/cooling-scenarios_1.14.2.pdf
>>> (bad)
>>> http://gyvern.ipc.hiroshima-u.ac.jp/~mpsuzuki/cooling-scenarios_1.14.3.pdf
>>> (good)
>>> At present, I'm not sure if 1.14.3 should be released
>>> as it is in git head, or, 1.14.3 should prioritize older
>>> Mac OS X functions as far as they are available.
>>> If you have your comment, please post...
>>>
>>> Regards,
>>> mpsuzuki
>>>
>>> suzuki toshiya wrote:
>>>> I found that my latest Mac OS X (10.6 :-)) is capable
>>>> to check the new API, however, the bundled python is
>>>> too old (and i386) to work with homebrew cairo (most
>>>> developer tools are already designed for x86_64). I
>>>> have to start from homebrewing of Python, so it will
>>>> take longer than I expected... I'm sorry for that I
>>>> could not work in prompt time scale.
>>>>
>>>> Regards,
>>>> mpsuzuki
>>>>
>>>> suzuki toshiya wrote:
>>>>> Thanks!
>>>>>
>>>>> I will try to reproduce the issue.
>>>>>
>>>>> Regards,
>>>>> mpsuzuki
>>>>>
>>>>> Daven Quinn wrote:
>>>>>> OK will do.
>>>>>>
>>>>>> If you want to give this a shot, the svg is attached...
>>>>>>
>>>>>> Daven
>>>>>>
>>>>>> On Sat, May 9, 2015 at 9:35 PM, suzuki toshiya
>>>>>> <mpsuzuki at hiroshima-u.ac.jp>
>>>>>> wrote:
>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> Although I've not tested for the reproduction of the issue,
>>>>>>> the problem you reported reminds me a regression in 1.14.2
>>>>>>> which is resolved in the latest git head (fixed on 2015-04-24).
>>>>>>> The discussion is here (see the message posted on 2015-03-16,
>>>>>>> from Inkscape developer).
>>>>>>>         https://bugs.freedesktop.org/show_bug.cgi?id=84324
>>>>>>> Unfortunately, there are no screenshots, I don't know how
>>>>>>> the regression appear on the Inkscape.
>>>>>>>
>>>>>>> Could you try the latest git head to assure if it is yet
>>>>>>> another regression?
>>>>>>>
>>>>>>> Regards,
>>>>>>> mpsuzuki
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> suzuki toshiya wrote:
>>>>>>>
>>>>>>>> Hi,
>>>>>>>>
>>>>>>>> Because there was a discussion whether an obsolete Mac OS X
>>>>>>>> function to extract an outline data from a glyph in font
>>>>>>>> should be kept (to support legacy platform) or replaced by
>>>>>>>> newer function (for better maintenancability), I'm much
>>>>>>>> interested in this issue (although my hands are very slow).
>>>>>>>> I want to try to reproduce the problem. Could you post the
>>>>>>>> SVG that you used?
>>>>>>>>
>>>>>>>> Regards,
>>>>>>>> mpsuzuki
>>>>>>>>
>>>>>>>> Daven Quinn wrote:
>>>>>>>>
>>>>>>>>> Hello,
>>>>>>>>>
>>>>>>>>> I recently upgraded to Cairo 1.14.2 (installed via Homebrew) on my
>>>>>>>>> Mac
>>>>>>>>> OS X
>>>>>>>>> (latest) system. I am using it via CairoSVG to create PDF
>>>>>>>>> graphics. The
>>>>>>>>> upgrade broke the glyph rendering of CairoSVG, resulting in grossly
>>>>>>>>> too-large font glyphs (with the same baseline/spacing as prior,
>>>>>>>>> leading
>>>>>>>>> to
>>>>>>>>> compressed text). I have tried building from source and relinking.
>>> An
>>>>>>>>> example of the output is shown below.
>>>>>>>>>
>>>>>>>>> Downgrading to cairo 1.12.16 fixes the issue (see attached),
>>>>>>>>> although the
>>>>>>>>> following warning is raised (which does not appear in 1.14):
>>>>>>>>>
>>>>>>>>> May  4 18:37:47 Daven-Quinn python2.7[993] <Error>: The function
>>>>>>>>> ‘CGFontGetGlyphPaths’ is obsolete and will be removed in an upcoming
>>>>>>>>> update. Unfortunately, this application, or a library it uses, is
>>>>>>>>> using
>>>>>>>>> this obsolete function, and is thereby contributing to an overall
>>>>>>>>> degradation of system performance.
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Any insights on a possible fix?
>>>>>>>>>
>>>>>>>>> Daven
>>>>>>>>>
>>>>>>>>> note: I tried sending with the below-linked files as attachments,
>>>>>>>>> but it
>>>>>>>>> was over the 150 kB (!) limit for message size, so I'm not sure if
>>>>>>>>> the
>>>>>>>>> email got through. A graphics library should accept issues in
>>>>>>>>> graphical
>>>>>>>>> form! It seems like source-control should be migrated to a newer
>>>>>>>>> platform
>>>>>>>>> to ease community contributions.
>>>>>>>>>
>>>>>>>>>>>>>>>>>>  cairo-1.14.pdf
>>>>>>>>> <
>>>>>>>>>
>>> https://docs.google.com/a/davenquinn.com/file/d/0BzFTyjxorIWAZ042NUJzVGNEZl8waktVRXpZRWZLaVZuVEJN/edit?usp=drive_web
>>>>>>>>>
>>>>>>>>> ​​
>>>>>>>>>  cairo-1.12.pdf
>>>>>>>>> <
>>>>>>>>>
>>> https://docs.google.com/a/davenquinn.com/file/d/0BzFTyjxorIWAYnExd2JfQWU3ZzZSdjBDU3VDa21WOExfOC0w/edit?usp=drive_web
>>>>>>>>>
>>>>>>>>>>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
> 
>> -- 
>> cairo mailing list
>> cairo at cairographics.org
>> http://lists.cairographics.org/mailman/listinfo/cairo
> 


More information about the cairo mailing list