Hi Jack:

The ccp4 package currently works and compiles the way that users  
expect it to
on OS X, on linux, and 10 other platforms.  If we make large changes,  
chances are
users simply won't want to use this any more.  Also,  I just don't  
have time right now
to re-invent the whole compilation scheme. Every build takes hours.   
I've already invested
countless hours in this, along with ccp4, to make it work on OS X,  
and as of today it works
great, I can get results and publish them in the best journals. CCP4  
have a paid staff to
address compilation issues that arise, courtesy of the British  
taxpayers.  They also now
distribute OS X binaries.  It just doesn't make sense for us to  
invest more time in this. It
would probably make far more sense for me just to remove ccp4 and all  
of the other
packages that don't compile with your new gfortran from fink, or else  
keep the current
gcc4 package and call it something like gcc4-old.

Bill



On Nov 15, 2006, at 9:02 AM, [EMAIL PROTECTED] wrote:

> Bill,
>    No problem. If you have ever followed the debian packaging process,
> they spend a huge amount of time detecting and eliminating unexpected
> library dependencies. Of course they have a major advantage since they
> use build machines for something like 11 architectures. So at any  
> given
> time each of those build machines has some random set of packages
> installed that might tickle latent build issues.
>    It will be interesting to find out if you sense any speed up from
> the use of -O3. Their use of -O0 and -O1 for whole directories was
> pretty severe overkill for the problem. What I would do is find each
> program that fails at -O3 and try -O1 for all of its files and then
> subsets of -O1 and -O3 until I pinned down the offending file. At that
> point we can try to puzzle out a testcase.
>   I did notice that the ccp4 package is pretty fat even with shared  
> libs.
> I need to review the total build log and see if there is any other
> places where we can convert the build over to shared libs. It can make
> a huge difference. You should have seen Mklinux when we did the first
> release before they had shared lib support. Bloated binaries  
> everywhere.
>                      Jack
>
> ----- Original Message -----
> From: William Scott <[EMAIL PROTECTED]>
> Date: Wednesday, November 15, 2006 11:45 am
> Subject: Re: full multilib packaging
> To: [EMAIL PROTECTED]
>
>> Hi Jack:
>>
>> I haven't had time to do anything more than compile ccp4 with the
>> old
>> and new gfortran.  (Sorry, my wife is gone for 2 weeks and I have 3
>>
>> little kids that are relentless in their demands).
>>
>> The phaser issue should be easy enough to correct, since it works
>> in
>> isolation of everything else.  Phaser is really more of a fellow
>> traveller with ccp4 than an integrated part. There is actually a
>> reason to keep it the way it is.  I have a separate cctbx package
>> that builds similar libraries, etc.  that uses the system
>> compilers,
>> system python, and so on.  So I do not install these libraries from
>>
>> the ccp4 build, so it is probably easiest just to leave things as
>> they are in that case.
>>
>> Refmac is the ccp4 refinement program.  The newer version works
>> with
>> -03.  I have a separate package that supercedes what is in ccp4, so
>>
>> it probably isn't worth investing time in an obsolete version of
>> refmac.  Most of the other programs aren't time-intensive, except
>> for
>> the molecular replacement programs (molrep, amore) and density
>> modification (dm, solomon), so those are the most reasonable
>> priorities.
>> I haven't yet had a chance to try your additional patch, and wont
>> have before this evening.  Sorry!
>>
>> Bill
>>
>>
>>


-------------------------------------------------------------------------
Take Surveys. Earn Cash. Influence the Future of IT
Join SourceForge.net's Techsay panel and you'll get the chance to share your
opinions on IT & business topics through brief surveys - and earn cash
http://www.techsay.com/default.php?page=join.php&p=sourceforge&CID=DEVDEV
_______________________________________________
Fink-devel mailing list
Fink-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/fink-devel

Reply via email to