r747 - in /trunk/boinc-app-seti/debian: README.Debian README.Debian-source changelog rules
steffen_moeller at gmx.de
Wed Jan 2 20:52:15 UTC 2008
Frank S. Thomas wrote:
> On Saturday 17 November 2007 23:18, smoe-guest at users.alioth.debian.org wrote:
>> Author: smoe-guest
>> Date: Sat Nov 17 22:18:03 2007
>> New Revision: 747
>> --- trunk/boinc-app-seti/debian/README.Debian (added)
>> +++ trunk/boinc-app-seti/debian/README.Debian Sat Nov 17 22:18:03 2007
>> +For the SPARC architecture, we introduced a test that checks if the
>> +machine is an UltraSparc and if so add "-mcpu=v9" to the compiler flags.
>> +This way, the binary will no longer be executable on older machines
>> +if the build daemon was an UltraSparc. Users of those systems will be
>> +required to locally build the package.
> I'm a bit uneasy about this change. First of all shipping a binary in the
> official sparc.deb package that intentionally fails to run on some machines
> IMO justifies a bug of severity "important" or higher. Debian packages should
> work right out of the box, without the need to recompile them. It is also not
> really easy to find out if the package was built on an UltraSparc or not, and
> therefore if it will run on a particular sparc machine or not. And how is the
> user supposed to recognize that the package needs to be recompiled?
> I think it spares us some bug reports and our sparc users some hassle if we'd
> comment out these sparc specific flags. If one wants an optimized SETI
> application, one can always download the source package remove the comments
> from debian/rules and build it with the UltraSparc optimizations.
I completely agree with what you are saying. Frankly, I was just a bit
surprised about finding this change of mine submitted to svn :) I cannot
recall if I had informed you about it, my latest attempts with SETI on
SPARC from mid December also failed.... without the mcpu option.
More information about the pkg-boinc-devel