On 25 January 2016 at 13:57, Martin Michlmayr wrote:
| * Dirk Eddelbuettel <e...@debian.org> [2016-01-24 10:45]:
| > 
| > On 22 January 2016 at 14:26, Martin Michlmayr wrote:
| > | > We've seen similar breakage with R packages with ABI/API changes.
| > | 
| > | You were right.  I rebuilt r-cran-rcpp and libquantlib0-dev and now
| > | rquantlib builds fine.
| > | 
| > | I then did:
| > | 
| > |     apt-get install --reinstall libquantlib0-dev libquantlib0v5
| > | 
| > | built again and got the segfault again.
| > 
| > That's a bit weird.
| 
| Sorry, maybe I wasn't clear.  It's completely consistent with what you
| said.

You were because ...
 
| I recompiled r-cran-rcpp and libquantlib0-dev with GCC 6, instaled
| them and rquantlib built with GCC 6.

... that is as expected while ...
| 
| I then went back to r-cran-rcpp and libquantlib0-dev from the archive
| (built with GCC 5), built rquantlib with GCC 6 and it failed.  (This
| is just the original situation.)

... this is still weird.

But I appreciate the gcc 5 / gcc 6 mixing explanation. I had missed that.

I guess the general rule is to, if in doubt, rebuild with the same compiler.
OTOH all the g++-5 pain to set us up for a smoother transition.  Maybe we're
not there yet.

Dirk

| > Yes please change it to minor. Or reply if you want me to.
| 
| I'll do it.
| 
| -- 
| Martin Michlmayr
| Linux for HPE Helion, Hewlett Packard Enterprise

-- 
http://dirk.eddelbuettel.com | @eddelbuettel | e...@debian.org

Reply via email to