Hi Fotis,
On 08/03/2012 04:25 PM, Fotis Georgatos wrote:
Hi,
here's my regular TGIF email ;-)
(1) Skipping certain EasyBuild steps
On 26/07/2012 17:07, Fotis Georgatos wrote:> * Allow to skip certain steps
easily (we notice it is often required):
** skip_configure=true
** skip_make=true
** skip_make_installdir=true
That reduces LOC (lines of code) and hence human error. It's more easy, too.
Finally, we've arrived to a trully elegant way to override certain steps:
preconfigopts='/bin/true'
premakeopts='/bin/true'
It certainly helps to minimize legth of easyblocks or, eliminate them.
This will work, but it looks more like a hack to me than anything else.
A proper solution would be to create dedicated classes for the use cases
we have, like the Binary class we now have in our develop branch:
https://github.com/hpcugent/easybuild/blob/develop/easybuild/easyblocks/b/binary.py
(2) Contribution towards v0.9
This regards the list at:
https://github.com/hpcugent/easybuild/wiki/List-of-supported-software-packages
(2a)
Cedric has made some more progress, intermediate results are visible here:
https://github.com/fgeorgatos/easybuild/tree/master/experimental
(2b)
Some more simple easyconfig files are available; review/cleanup needed.
These ones appear to build fine (for goalf), but they are otherwise untested:
(to be precise, they do not seem fail in the 3 basic steps & give modulefile)
(also, the libraries need to be declared as such)
<snip>
That's really nice, we should look into getting the example easyconfigs
and accompanying easyblocks for these into our develop branch.
Can you guys fork from develop, add the files there (where they belong
in the tree structure) and issue manageable pull requests, i.e. a couple
of (related?) applications at the same time?
(3) Wishlist updates: select logfile location
OK, the wishlist has been amended with one worthy to check idea:
https://github.com/fgeorgatos/easybuild/wiki/Wishlist
"""
Add possibility to save the logfile somewhere even when failures happen (/tmp
is no good for queue submitted jobs)
"""
This is actually quite critical for me to speed up operations,
since I run them in batch mode (cluster jobs) and the logfiles
-with all the interesting error information- are scattered around.
(the interesting info, are actually the failures, which go in /tmp!!!)
Can we do something about it? Am I missing some other option here?
We should get tickets submitted for these into our bug tracker, so we
can look into them.
This afternoon, we'll discuss internally how we can set up a bug tracker
for EasyBuild that we can open up for everyone. That should speed up
getting these features implemented.
We'll try and look into this particular log file location request soon.
What we do now is make EasyBuild spit out (debug) logging into to stdout
using the -ld option, and then redirecting that to a file on shared
storage. That way, the output stays available.
We only have some very interesting feature being implemented by our
summer intern to very easily submit builds to a PBS server; this might
also be very useful for you.
(4) Compiling/Linking against boost and such libraries
While building Cufflinks, Cedric came across this issue:
http://stackoverflow.com/questions/11688893/decoding-gcc-specs-file-line
To summarize what I understand, there is a subtle difference in the gcc
spec strings and we wonder if there is some funny interaction between
Debian/Easybuild/gcc; can you supply the following command from your side?
$ gcc -dumpspecs|grep -A1 multilib:
# Most RHEL clones give: . !m64 !m32;64:../lib64 m64 !m32;32:../lib !m64 m32;
(check it character-per-character, it's hard exercise for the human reader)
This returns :
*multilib:
. !m64 !m32;64:../lib64 m64 !m32;32:../lib !m64 m32;
Seems to match that comment.
(5) Next steps
Finally, I think we reach a point that a mini-hackathon would be welcome;
Especially, to touch on matters like the following and integrate current work:
https://github.com/fgeorgatos/HPC-RFC/blob/master/0001/0001.md
https://github.com/fgeorgatos/easybuild/wiki/Wishlist
Yes, I think that's a good idea.
How will we organize this? Are you in the neighbourhood anytime soon, or
should we plan a trip to your lab?
regards,
Kenneth
