[OpenJDK 2D-Dev] [PATCH FOR REVIEW] fix for bug 8011693: Remove redundant fontconfig files
Jiri Vanek
jvanek at redhat.com
Thu May 30 14:57:36 UTC 2013
On 05/30/2013 03:29 PM, Andrew Hughes wrote:
> ----- Original Message -----
>>> None of these are touched any more in the current patch, as we don't build
>>> on
>>> these targets.
>>
>> Right, I noted that in the first sentence of my prior email.
>>
>>> we haven't tested on *BSD.
>>
>> fair enough but the difference is that there is no build
>> line that ever copies this file, so its un-used.
>>
>
> Yes, it was added in:
>
> changeset: 5117:d45bc4307996
> user: michaelm
> date: Tue Mar 06 20:34:38 2012 +0000
> summary: 7113349: Initial changeset for Macosx port to jdk
>
> That patch doesn't make any Makefile changes to add it in, but it may still be used
> by *BSD users as they have other patches not yet upstreamed. I'd rather they made
> the decision as to whether or not it should be removed.
Ok. So I have kept it in.
>
>>> From my review of the patch, I think the variables are left empty, not
>>> removed:
>>
>> Looking at
>> http://jvanek.fedorapeople.org/oracle/jdk8/webrevs/removedFontConfigFiles-linuxOnly/makefiles/GendataFontConfig.gmk.sdiff.html
>>
>> they seem to be removed.
>>
>> I was expecting to see
>> GENDATA_FONT_CONFIG_SRC_DIR := \
>> 38 $(JDK_TOPDIR)/src/solaris/classes/sun/awt/fontconfigs
>>
>> may be left in there and also that it left just
>>
>> GENDATA_FONT_CONFIG_SRC_FILES :=
>>
>> rather than have it removed completely
>>
>
Ok. I have added them as you have suggested
http://jvanek.fedorapeople.org/oracle/jdk8/webrevs/removedFontConfigFiles-linuxOnly_3/
> Ok I was thinking of the 7 makefile:
>
> -FONTCONFIGS_SRC = $(PLATFORM_SRC)/classes/sun/awt/fontconfigs
> -_FONTCONFIGS = \
> - fontconfig.properties \
> - fontconfig.SuSE.properties \
> - fontconfig.Ubuntu.properties \
> - fontconfig.Fedora.properties
> +FONTCONFIGS_SRC =
> +_FONTCONFIGS =
> +
>
> When are these finally going to be removed? It's very confusing.
I'm afraid we can not just take them off as they are widely used.
I think empty is safer here then remove, as they are kept for mac and win port.
>
>> -phil.
>>
>>
>> On 5/29/2013 1:17 PM, Andrew Hughes wrote:
>>> ----- Original Message -----
>>>> Jiri,
>>>>
>>>> I think this has mostly been hashed out as the fix is reduced to Linux
>>>> but here's my over-due input :
I agree it have hashed, but fix the underlying windows issue is quite complex task. At least it
deserves another changeset. And especially it deserves windows specialist:)
>>>>
Yuhuuu! This is an feedback I'm happy for :)
It explains so much.....
>>>> 1) Windows *absolutely* still needs fontconfig files.
>>>>
>>>> 2) Mac OS X doesn't obey what's there but that doesn't mean its
>>>> going to work when you just remove them.
Then it should be tested and removed rather then keeping undetermined code.
>>>>
>>>> 3) Solaris *does* want the Solaris one, even though it has the
>>>> same fontconfig platform support as Linux. This is for compatibility.
>>>>
>>> None of these are touched any more in the current patch, as we don't build
>>> on
>>> these targets.
>>>
>>>> 4) Linux does not need them *so long as* fontconfig is there and working.
>>>> OpenJDK on any Linux desktop should be fine.
>>>>
>>>> 5) However policy decisions were made to leave them there for some
>>>> particular linux variants in the closed repo, again for compatibility,
>>>> but you are leaving that alone, so that's fine, although we should
>>>> revisit this
>>>>
>>>> 6) I never noticed the bsd one before. It must have snuck in with mac
>>>> port.
>>>> Is anything even referencing it ? If not I think this can be removed too.
>>>>
>>> I don't think this should be part of this patch for the same reason
>>> Solaris/Mac OS/Windows
>>> aren't; we haven't tested on *BSD.
agree
>>>
>>>> 7) The files themselves and support for the files are distinct issues.
>>>> There should still be the ability for [say] Gentoo, to decide they want
>>>> a particular set of fonts used and so they will ship a file .
>>>> So it might be better to leave the variables there (empty) and with
>>>> a comment that this is a placeholder.
>>> We do actually carry one for both Gentoo and RHEL in IcedTea for OpenJDK 6.
>>> From my review of the patch, I think the variables are left empty, not
>>> removed:
>>>
>>> http://jvanek.fedorapeople.org/oracle/jdk8/webrevs/removedFontConfigFiles-linuxOnly/jdk.patch
>>>
>>>> -phil.
>>>>
>>>> On 5/29/2013 5:06 AM, Andrew Hughes wrote:
>>>>> ----- Original Message -----
>>>>>> On 05/20/2013 04:37 PM, Jiri Vanek wrote:
>>>>>>> On 05/10/2013 04:08 PM, Jiri Vanek wrote:
>>>>>>>> On 04/08/2013 05:31 PM, Jiri Vanek wrote:
>>>>>>>>> On 04/08/2013 04:13 PM, Vladislav Karnaukhov wrote:
>>>>>>>>>> Hello Jiri,
>>>>>>>>>>
>>>>>>>>>> please see inline.
>>>>>>>>>>
>>>>>>>>>> On 4/8/2013 05:29 PM, Jiri Vanek wrote:
>>>>>>>>>>> On 04/08/2013 02:39 PM, Vladislav Karnaukhov wrote:
>>>>>>>>>>>
>>>>>>>>>>> Thank you very much for win-check! It will force me to install new
>>>>>>>>>>> windows machine somewhere.
>>>>>>>>>>> Do you mind do check if pure removal of fontconfig files (both src
>>>>>>>>>>> and
>>>>>>>>>>> bfc) from you installed jdk7/8 on windows will work? (should)
>>>>>>>>>> Yes, I've checked and it does *not* work. That's the reason why I
>>>>>>>>>> replied to your very first
>>>>>>>>>> message. A removal of fontconfig.* files simply crashes Java, - on
>>>>>>>>>> both
>>>>>>>>>> Windows and Mac, - because
>>>>>>>>>> some font management-related classes rely on these files. Hence my
>>>>>>>>>> question regarding deeper
>>>>>>>>>> re-design on font management system...
>>>>>>>>>>
>>>>>>>>>> I've tested Mac build as well, and there's the same error:
>>>>>>>>> Ok. I will try anyway:)
>>>>>>>>> For linux I'm quite sure the new fontmanagers are working pretty
>>>>>>>>> fine.
>>>>>>>>> Do you think it will be acceptable to prepare smaller clean up - to
>>>>>>>>> remove all linux fontconfig
>>>>>>>>> files?
>>>>>>>>>
>>>>>>>>> And later, as separate changeset to fontmanagers for windows/mac,
>>>>>>>>> but
>>>>>>>>> I'm afraid I will not be
>>>>>>>>> capable of such an development on non linux system.
>>>>>>>>>
>>>>>>>>> Thanx for your help,
>>>>>>>>>
>>>>>>>>> J.
>>>>>>>> Hi!
>>>>>>>>
>>>>>>>> I had finally found some free time, so here it is - smaller version
>>>>>>>> which
>>>>>>>> is removing just stuff for
>>>>>>>> linux when OpenJDK is defined.
>>>>>>>>
>>>>>>>> http://jvanek.fedorapeople.org/oracle/jdk8/webrevs/removedFontConfigFiles-linuxOnly/
>>>>>>>>
>>>>>>>> Although I had windows build, I lost this machine so - again (and
>>>>>>>> sorry
>>>>>>>> for that) - tested only on
>>>>>>>> Fedora.
>>>>>>>>
>>>>>>>> Also when I read the individual fontmanagers, I believe that they
>>>>>>>> really
>>>>>>>> *should* work without
>>>>>>>> fontocfigs. So although this is fixing the 8011693, new bugs should be
>>>>>>>> filled for windows and mac,
>>>>>>>> because theirs implementations are broken.
>>>>>>>>
>>>>>>>> Thank you very much for any comments.
>>>>>>>>
>>>>>>>> Best Regards
>>>>>>>> j.
>>>>>> Ping?
>>>>>>
>>>>>> Any advice how to move this forward?
>>>>>>
>>>>>>> I know that this is minor fix compared to others I can read on this
>>>>>>> channel, but as the font
>>>>>>> managers exists, and fontconfig files *should* be redundant, then this
>>>>>>> change should be done. If
>>>>>>> fontmanagers are buggy (and eg windows one appeared to be) then as soon
>>>>>>> as
>>>>>>> this will be tempted then
>>>>>>> sooner it will get fixed.
>>>>>>>
>>>>>>> For linux I'm pretty sure this is working, and we have even removed the
>>>>>>> fontconfig files from
>>>>>>> packages in public facing version three months ago [1]
>>>>>>>
>>>>>>> So this can be first step to get rid of old and redundant font mapping
>>>>>>> completely.
>>>>>>>
>>>>>>>
>>>>>>> J.
>>>>>>>
>>>>>>>
>>>>>>> [1]
>>>>>>> http://pkgs.fedoraproject.org/cgit/java-1.7.0-openjdk.git/commit/?h=f17&id=9d6dd62ae2123635b4d15e40e527a0b617756484
>>>>>>>
>>>>>>> (search for +rm %{buildoutputdir}/j2re-image/lib/fontconfig)
>>>>> I've applied this patch and built OpenJDK, and it went fine. A basic
>>>>> Swing
>>>>> application still loads up fine.
Thank you very much for review!
I was already lost in despair that this rotten files will remain inside.
>>>>>
>>>>> So looks good to go to me.
>>>>
>>
>>
>
More information about the 2d-dev
mailing list