On Nov 24, 2010, at 4:31 AM, Pavan Balaji wrote:
> 1. Gracefully abort at configure time if the compiler doesn't support C99: https://trac.mcs.anl.gov/projects/mpich2/changeset/7497/mpich2/trunk/src/pm/hydra/tools/bind/hwloc/hwloc
(now that I'm on a proper keyboard...)
I'm ok with this, even if it's redundant elsewhere. Defensive programming and all that.
> 2. Slight improvement to the error message: https://trac.mcs.anl.gov/projects/mpich2/changeset/7499/mpich2/trunk/src/pm/hydra/tools/bind/hwloc/hwloc
No problem.
> 3. Warning squashes from shadowing global variables: https://trac.mcs.anl.gov/projects/mpich2/changeset/7501/mpich2/trunk/src/pm/hydra/tools/bind/hwloc/hwloc
I have no argument with this one, either.
> The following patches are just a matter of preference. So you might not want to incorporate them, but here they are anyway:
>
> 1. Disable doxygen builds by default: https://trac.mcs.anl.gov/projects/mpich2/changeset/7498/mpich2/trunk/src/pm/hydra/tools/bind/hwloc/hwloc
I agree with Brice on this one; you should be using embedded mode to disable a bunch of stuff by default. This will save you lots of hassles; not just doxygen.
So my question -- per prior email -- is: why not use --disable-embedded-mode, and --disable-shared --enable-static? (vs. patching) IIRC, I put --disable-embedded-mode there exactly for the case of projects that want to embed but don't want to call our m4 directly.
> 2. Don't automatically pick debugging or picky compiler options unless the user asked for them: https://trac.mcs.anl.gov/projects/mpich2/changeset/7500/mpich2/trunk/src/pm/hydra/tools/bind/hwloc/hwloc
I originally asked for this policy (do debug builds by default when you find .svn), but I could probably be convinced to reverse it pretty easily. So -- I'm ok with it.
--
Jeff Squyres
jsquyres_at_[hidden]
For corporate legal information go to:
http://www.cisco.com/web/about/doing_business/legal/cri/
|
![[Flag]](../../../../../community/mirrors/flag-us.gif){kind=small}