On 6 November 2010 06:43, Mitesh Patel <qed...@gmail.com> wrote:
> On 11/05/2010 07:25 PM, Jason Grout wrote:
>> On 11/5/10 5:16 PM, Mitesh Patel wrote:
>>> On 11/05/2010 07:49 AM, kcrisman wrote:
>>>>
>>>>> http://sage.math.washington.edu/home/buildbot/binaries/
>>>>>
>>>>> Does it usually take a long time (~1 hour) to make a dmg
>>>>> distribution on
>>>>> bsd.math, or more generally, on OS X?
>>>>
>>>> It can take quite a while on a slower machine; I feel like bsd.math is
>>>> fairly new, so it shouldn't take quite that long, but on a similar
>>>> order of magnitude (i.e., not under 10 minutes).  Is bsd.math pretty
>>>> busy with other things?
>>>
>>> I checked a few times during the process and did not see any major
>>> other activity.  However, it took only about 20 minutes when I tried
>>> again.
>>>
>>> On OS X, the output from 'sage -bdist' includes
>>>
>>> If you wish to create a Mac Application please set
>>> SAGE_APP_BUNDLE=yes
>>> Creating dmg
>>> (If you don't wish to create a disk image please set SAGE_APP_DMG=no)
>>>
>>> Should I have the builder make an Application dmg, too?  Is
>>
>> That would be fantastic.
>>
>>> sage-4.6.1.alpha0-OSX-64bit-10.6-i386-Darwin-app.dmg
>>>
>>> a good canonical name [1] for this, given that the default canonical
>>> name on bsd is
>>>
>>> sage-4.6.1.alpha0-OSX-64bit-10.6-i386-Darwin.dmg
>>
>> Sure!
>
> Done.  The latest run took over 2 hrs 20 mins to finish consecutive
> 'sage -bdist' commands on an otherwise idle bsd.math.  Strange.

I think that sage-bdist could be speeded up.

At one point a huge tar file is created then extracted in another
directory to perform a copy. That takes a long time. If instead the
"copy" was done by piping stdout of one tar process into stdin of
another, it would save creating the intermediate file. That would save
a lot of totally unnecessary I/O. Copying in out out of memory (which
is what a pipe does), will use less I/O then reading and writing to
disk.

If the unportable 'cp -a' was replaced with another tar to tar copy,
the Solaris issues could be resolved.


There are other things that could be useful in that script too - the
ability to use p7zip / lzma compression instead of bzip2. While that
is a LOT more CPU intensive to create the file,

There are a lot of things that could be improved in that script, but
of course they all take time to implement.

Dave

-- 
To post to this group, send an email to sage-devel@googlegroups.com
To unsubscribe from this group, send an email to 
sage-devel+unsubscr...@googlegroups.com
For more options, visit this group at http://groups.google.com/group/sage-devel
URL: http://www.sagemath.org

Reply via email to