[OpenJDK 2D-Dev] RFR(XS): 8132408 : Check os.name before os.version in SunGraphicsEnvironment constructor

Phil Race philip.race at oracle.com
Tue Jul 28 19:08:09 UTC 2015


OK I'll accept this patch since I've dug around and found we have
one other place (in FontUtilities) which independently sets a different 
variable of
the same name only when uname -s (yes, I meant -s when I wrote -n :-)),
reports SunOS. And additionally the only place where the one you are
setting does anything meaningful is also guarded by an indirect check
of the same. I am not sure how this all came about and clearly
there needs to be some more tricky follow-up work, but your
patch will make things consistent.

-phil.

On 07/28/2015 11:43 AM, Volker Simonis wrote:
> Hi Phil,
>
> thanks for looking at this change.
> Please find my comments inline:
>
> On Tue, Jul 28, 2015 at 7:12 PM, Phil Race <philip.race at oracle.com> wrote:
>> The idea behind the check dates from when there was a big push
>> behing OpenSolaris and 'uname -n' would not necessarily return "SunOS"
>> but the version numbers shhuld match for an opensolaris built from the same
>> sources.
>>
>> So restricting the check to when the os.name is SunOS basically
>> breaks the whole point.
> I had the same fear initially, but then I found this:
>
> http://wiki.illumos.org/display/illumos/Modernizing+Uname
>
> entry where the Illumos people started a discussion about changing the
> output from 'unmae -s' to something different than "SunOS" in the
> future. But it seems they've still not changed that, so maybe this fix
> is 'good enough' for now?
>
> What else would you suggest otherwise?
>
> We have people who run Java in some highly secure environments and
> they are complaining about the VM trying to open this non-existing
> file.
>
>> But OpenSolaris is no more (isn't it?)
>> Probably the thing to do is also whack the parsing of /etc/release and just
>> keep the check for whether the Courier New font is there.
>> Or maybe for > 5.10 we should just always use fontconfig like we do on
>> Linux nowadays ?
>>
> I'm by no means an OpenSolaris expert (and I don't have a machine to
> check) but I will be happy to add whatever additional change you
> propose.
>
> Regards,
> Volker
>
>> -phil.
>>
>>
>> On 07/28/2015 09:00 AM, Volker Simonis wrote:
>>> Hi,
>>>
>>> could you please review the following small change:
>>>
>>> http://cr.openjdk.java.net/~simonis/webrevs/2015/8132408/
>>> https://bugs.openjdk.java.net/browse/JDK-8132408
>>>
>>> In the constructor of SunGraphicsEnvironment there's a check for
>>> os.version > 5.10.
>>>
>>> If this check is true, /etc/release is opened to get additional
>>> Solaris-specific information. But this query only makes sense on
>>> Solaris, so we should first check if we're really running on Solaris
>>> before we check for os.version being bigger than 5.10.
>>>
>>> Otherwise, the VM will try to open /etc/release on every system which
>>> has an os.version bigger than 5.10 (e.g. AIX 7.1 or Linux once we get
>>> to kernel version 5.11 :).
>>>
>>> Thanks,
>>> Volker
>>




More information about the 2d-dev mailing list