Hello Jacek and Daniel,
Daniel has a very good analysis of the new system. Here's some
additional informations :
*
**The install procedure completely support eups*. And yes, you've got to
customize it if you don't rely on eups (using custom.py should be enough).
Today, i've updated the documentation for the master branch (see
README.txt), and packaged a Qserv master release on in2p3 distserver.
The README.txt procedure described in master branch now installs Qserv
master release.
Then you have to read *README-devel.txt* if you want to enable your
Qserv current version (i.e. the one in your git repos) in the eups
stack. It's quite straightforward once you've merge the new build system
(i.e. top-level SConstruct file) and you've done it a first time.
Please let me know if it works, on my machine it was fine.
I propose you to work on this procedure during the hackaton :
- i could first show you this way of developing using eups and Qserv,
- with Mario, we could then define a better packaging and developing
procedure relying on LSST standards tools for continuous integration
P.S. :
- Daniel, please note that i had to undo your commit n°
63c51c29abfee5bc5573cc23413f8867f3b9d490, to make the system work with eups.
- Please note that Ubuntu 12.04 isn't supported (due to a antlr version
complicated to set up), but only 13.10 (whereas i haven't tested this
very las packaging on it, dur to time constraints).
Note also that next Ubuntu LTS version (14.04), is coming by the end of
april. I think this last one may be easier to support than 12.04.
See you soon,
Fabrice
On 04/17/2014 03:15 AM, Daniel L. Wang wrote:
> Jacek,
>
> I had to fight a lot with the new system.
>
> (Fabrice, I really appreciate what you've done. You made a lot of
> fixes, and there are a lot of things that are cleaner. But there are a
> couple of things that can be improved.)
>
> Basically, you need to run scons from the top level (qserv/, not
> qserv/core). There are some assumptions that an EUPS-like structure
> exists that you'll need to workaround (hopefully, we can isolate these
> assumptions into one place so the EUPS-adaptations can be switched
> off/on).
>
> * You'll need a custom.py. It is not optional (we should change it so
> that it is optional as much as possible). Start with the
> example.custom.py file. You don't need it to be absolutely correct,
> because after a couple basic checks, the system doesn't really use all
> of the entries.
> * You'll also need to put the dependencies into your path. I isolate
> this by having a build.sh script that alters PATH before calling
> scons. This is not quite desirable.
> * There are some assumptions that break if the dependencies are
> system-installed in certain cases, but hold if the environment is
> EUPS-managed.
> * The dist/ directory is only created if you ask for "scons install",
> not if you just say "scons".
>
> I made a couple patches as well that I snuck into DM-282 before I
> merged (make the system more accepting of non-eups systems, fix a
> broken(?) path for the SWIG output). You probably want them.
>
> If your dev approach is like mine, I think this is all you need to
> worry about. I don't know much about the admin scripts (scripted
> compilation, daemon startup/shutdown).
>
> Hope this helps,
> -Daniel
>
>
> On 04/16/2014 05:48 PM, Jacek Becla wrote:
>> Daniel: how did you build/test qserv when you were merging
>> your code in? Given I can't build qserv, I am completely
>> stuck with rebasing my branches and getting ready for the
>> final merge
>>
>> Jacek
>>
>> p.s. in the email below the
>>
>> "But the core/SConstruct is not gone"
>>
>> should of course be
>>
>> "But the core/SConstruct is *now* gone"
>>
>>
>> Jacek
>>
>>
>>
>>
>>
>> On 04/16/2014 05:34 PM, Jacek Becla wrote:
>>> This is probably a question for Fabrice
>>>
>>>
>>> I have all dependencies setup on my laptop, and to build
>>> qserv I just run scons from core directory. But the
>>> core/SConstruct is not gone:
>>>
>>> commit 259f40dd89b03a3f93cfd14c5e4c493780e67daf
>>> Author: Fabrice Jammes <[log in to unmask]>
>>> Date: Mon Feb 24 17:55:33 2014 +0100
>>> - core/Sconstruct replaced by a top-level Sconstruct
>>>
>>>
>>> so... how do I build Qserv now??? (preferably on ubuntu 12.04)
>>>
>>>
>>> I looked at README.txt...
>>>
>>> it suggests to run
>>>
>>> admin/bootstrap/qserv-install-*
>>>
>>> I don't think I need it, I have a working version of
>>> qserv, (it is just based on different branch). Plus,
>>> there is no version for ubuntu12.04. So I skipped that.
>>>
>>> then README.txt suggests:
>>>
>>> INSTALL_DIR=/opt/example-qserv-install/
>>> eupspkg/newinstall-qserv.sh -H ${INSTALL_DIR}
>>>
>>> but I am immediately getting
>>>
>>> bash: eupspkg/newinstall-qserv.sh: No such file or directory
>>>
>>> I can only see newinstall-qserv-template.sh
>>>
>>> am I supposed to modify this template file somehow?
>>>
>>> Jacek
>>>
>>> ########################################################################
>>>
>>> Use REPLY-ALL to reply to list
>>>
>>> To unsubscribe from the QSERV-L list, click the following link:
>>> https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=QSERV-L&A=1
>>>
>> ########################################################################
>> Use REPLY-ALL to reply to list
>>
>> To unsubscribe from the QSERV-L list, click the following link:
>> https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=QSERV-L&A=1
>
> ########################################################################
> Use REPLY-ALL to reply to list
>
> To unsubscribe from the QSERV-L list, click the following link:
> https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=QSERV-L&A=1
########################################################################
Use REPLY-ALL to reply to list
To unsubscribe from the QSERV-L list, click the following link:
https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=QSERV-L&A=1
