encumbrances update

Kelly O'Hair Kelly.Ohair at Sun.COM
Fri Aug 10 01:02:49 UTC 2007

Ted Neward wrote:
> Well... several thoughts come to mind:
> (1) The fact that OpenJDK was not aiming to be buildable on Windows from the
> beginning was definitely NOT clear, at least not to me.
> (2) The fact that it would be viewed as acceptable to release the source in
> an unbuildable form on any of the "supported" platforms--Windows, Linux,
> Solaris--surprises me. I accepted the idea that it was unbuildable as I
> thought it was a temporary break, to be fixed "real soon now", not an
> acceptable state of affairs.
> (3) The idea that Windows is not as important as Linux or Solaris is a
> dangerous idea, IMHO. It essentially suggests that Sun is throwing away a
> whole legion of developers who work on Windows and would want to contribute
> patches and suggestions for improving the Java-Windows experience. IMHO, the
> success of .NET due in no small part to the fairly substantive gap between
> Java and Windows.
> (4) I don't want to suggest in any way that I'm unhappy with the progress
> made by you (Kelly) or any of the other build team developers--having
> wrestled with other build systems in the past, I'm stunned at the size of
> building the JVM, and awed at the fact that it works at all, much less on
> systems that aren't tightly locked down in terms of tools and filesystem
> layout. (I worked at Intuit for a while, on Quicken 5, and there you got to
> choose what drive letter the code would be installed on--several guys had
> external SCSI drives they took with them to work from home--but beyond that,
> everything had to look "just like this".) I said it publicly on my blog, I
> said it to Simon Phipps during an interview with him at OSCon (which is
> going up on iTunes), I'll say it again here: you guys are doing a DAMN
> impressive job.
> (5) It's kinda all moot--I did a fresh re-fetch from SVN for the b17 drop,
> used Ivan's build of the freetype DLL, and it all builds, both debug and
> fastdebug.


My sincere apologies for items 1-3 above. I'll give myself 20 lashes. :^(


> So yes, Dan, somebody outside of Sun managed to make it work on Windows. :-)
> Ted Neward
> Java, .NET, XML Services
> Consulting, Teaching, Speaking, Writing
> http://www.tedneward.com
>> -----Original Message-----
>> From: build-dev-bounces at openjdk.java.net [mailto:build-dev-
>> bounces at openjdk.java.net] On Behalf Of Kelly O'Hair
>> Sent: Thursday, August 09, 2007 10:22 AM
>> To: Dan Fabulich
>> Cc: discuss at openjdk.java.net; Anthony Petrov; Phil Race; Igor
>> Nekrestyanov; build-dev at openjdk.java.net
>> Subject: Re: encumbrances update
>> I thought it was clear from the beginning, perhaps I wasn't making this
>> as well known as I should. If so, my aplogies.
>> It was never expected that the initial OpenJDK source drops would be
>> buildable
>> on Windows. We focused on Linux and OpenSolaris and purposely left out
>> Windows due to the t2k issue, and knowing that the team working on this
>> was very close to removing it as a dependence, we felt this was
>> justfified.
>> So that fact that it has never built on Windows was well known, I
>> thought.
>> While I was trying to make t2k.lib available in the binary plugs,
>> another team
>> was busy trying to remove it completely. So you can't say we haven't
>> been trying
>> to deal with this issue.
>> It is pretty impossible to guarantee builds will be successful in all
>> possible configurations, but once we get past the basic issues, we will
>> do regular test builds of OpenJDK on Windows. But speaking from years
>> of
>> experience, Windows is a difficult platform to get consistent and
>> reliable.
>> Just because we can build on Windows isn't worth as much as saying it
>> for
>> Linux or OpenSolaris.
>> -kto
>> Dan Fabulich wrote:
>>> Igor Nekrestyanov wrote:
>>>> Of course i've tested these changes on all platforms including both
>> 32
>>>> and 64 bit Windows. It is "believed to build and work" because we
>> had
>>>> not performed full testing of openjdk binaries and we know that
>> build
>>>> can be fragile due to different build environments.
>>> I'm sure you've tested your changes in your build, and maybe that's
>> all
>>> I can ask for, but I'm pretty sure the same could have been said of
>> all
>>> of the previous builds, including the code drop we got in May and
>> every
>>> build since, all of which, we know, don't work for anyone who isn't
>>> @sun.com (due to at least one missing file).
>>>> My tests are not "ideal" for number of reasons:
>>>> 2) I was using binary plugs created from my personal workspace.
>>>>      I believe they should be the same as those to be published with
>>>> b17 code drop but this is my assumption.
>>> We know for certain that this assumption is FALSE.  No binary plug
>> given
>>> to the public has ever built successfully on Windows; since your
>>> personal plugs have been working for you for months, we can conclude
>>> that there's something critically different between your plugs and
>> the
>>> plugs we get.
>>> But who knows?  Maybe we'll get lucky this time.  :-)
>>>> BTW, for windows build you will need to build freetype.dll (i tried
>>>> only dll built with visualc). I believe that freetype build system
>>>> does not generate .dll on windows and therefore some manual tweaking
>>>> of freetype makefiles (and sources) might be necessary.
>>> Sounds like fun! :-)  Will the documentation include a patch?  Or
>> just
>>> some tips on how to make a .DLL?  What source files needed to change?
>>> -Dan
>> No virus found in this incoming message.
>> Checked by AVG Free Edition.
>> Version: 7.5.476 / Virus Database: 269.11.11/944 - Release Date:
>> 8/9/2007 2:44 PM
> No virus found in this outgoing message.
> Checked by AVG Free Edition. 
> Version: 7.5.476 / Virus Database: 269.11.11/944 - Release Date: 8/9/2007
> 2:44 PM

More information about the discuss mailing list