Skip to Content.
Sympa Menu

charm - Re: [charm] [ppl] megatest fails for ibverbs and --with-production option (charm-6.4.0)

charm AT lists.cs.illinois.edu

Subject: Charm++ parallel programming system

List archive

Re: [charm] [ppl] megatest fails for ibverbs and --with-production option (charm-6.4.0)


Chronological Thread 
  • From: Phil Miller <mille121 AT illinois.edu>
  • To: Gengbin Zheng <gzheng AT illinois.edu>
  • Cc: Thomas Albers <talbers AT binghamton.edu>, "charm AT cs.uiuc.edu" <charm AT cs.uiuc.edu>
  • Subject: Re: [charm] [ppl] megatest fails for ibverbs and --with-production option (charm-6.4.0)
  • Date: Sat, 31 Mar 2012 10:27:35 -0500
  • List-archive: <http://lists.cs.uiuc.edu/pipermail/charm>
  • List-id: CHARM parallel programming system <charm.cs.uiuc.edu>

On Sat, Mar 31, 2012 at 10:18, Gengbin Zheng
<gzheng AT illinois.edu>
wrote:
> Good to hear that. Looks like our fighting against the stack smashing
> protection continues. But I am a little surprised that syscall
> swapcontext can fail too with the same error.
> We should then include this flag in the current version. Or figure out
> a way to check/test longjmp error or at configure time to decide such
> a flag is needed or not.

In this case, it makes more sense to just add the flag
unconditionally. It will be meaningless on gcc versions that don't
enable Fortify, but it won't cause errors.

> Our nightly build on local machines are on gcc 4.4.3, and so 4.5 and
> later and not officially tested.

Not by autobuild perhaps, but my home machine has used the newer
compilers as soon as they're out - they're getting tested quite
regularly.

> We should keep some of our machines in lab updated to the latest
> version of OS and compilers for testing purpose.

Probably not a bad idea. However, in this case, it wouldn't have
helped, since Fortify isn't a new/old thing. It's a non-default OS
configuration that some Gentoo installations enable, and Debian and
Ubuntu haven't.

Regardless, I'll see if I can reproduce the failure and then ensure
that it stays away with the flag set.

Phil




Archive powered by MHonArc 2.6.16.

Top of Page