Sean argues the point well. I would add that having a special set of flags for one file should be done as a last resort (e.g., as a temporary work around a compiler bug, in which case a reference to the bug report should be provided). In other cases (e.g., tests like "size > size+1" replaced by "size == INT_MAX") it is better to rely on a defined constant than some potentially undefined behaviour.
In this spirit, would it be better to #define the required values (and perhaps provide checks to let us know when a value is incorrect)? Perhaps the existing code could provide a stand-alone tool to help compute values for new compilers/platforms where RTFM or contacting the tool developers fails. In the long run, perhaps these values would be provided with the compilers' headers. On Thu, Oct 17, 2013 at 4:47 PM, Raymond Lu <[email protected]> wrote: > Sean, > > Several of us including Quincey and Elena looked at the issue. We decided > that since the algorithm is trying to detect the alignment of integers, > ideally the flag -fcatch-undefined-behavior and other optimization flags > shouldn't to be used for H5detect.c. In the future, we can separate flags > for H5detect.c from the rest of the library. > > (For those who don't know what Issue 8147 is: CLANG compiler on mac > machines with the options -fcatch-undefined-behavior and -ftrapv catches > some undefined behavior in the alignment algorithm of the macro DETECT_I in > H5detect.c.) > > Please let us know your thoughts. > > Thanks. > > Ray > > > On Oct 9, 2013, at 3:48 PM, Elena Pourmal wrote: > > Sean, > > Unfortunately the bug was not fixed, but its priority has been bumped > (clang is one of the compilers that we should fully support). > > We are aware of the failures you have been posting to our CDash and will > try to resolve the issue ASAP. > > Elena > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > Elena Pourmal The HDF Group http://hdfgroup.org > 1800 So. Oak St., Suite 203, Champaign IL 61820 > 217.531.6112 > ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > > > > > On Oct 9, 2013, at 1:34 PM, Sean McBride <[email protected]> > wrote: > > On Tue, 8 Oct 2013 20:15:17 -0500, Elena Pourmal said: > > A pre-release candidate of HDF5 1.8.12 is available for testing and can > be downloaded from the following location: > > *SNIP* > > The new release includes but is not limited to: > > > Is HDFFV-8147 fixed? > > Cheers, > > -- > ____________________________________________________________ > Sean McBride, B. Eng [email protected] > Rogue Research www.rogue-research.com > Mac Software Developer Montréal, Québec, Canada > > > > _______________________________________________ > Hdf-forum is for HDF software users discussion. > [email protected] > > http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org > > > > _______________________________________________ > Hdf-forum is for HDF software users discussion. > [email protected] > > http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org > > -- George N. White III <[email protected]> Head of St. Margarets Bay, Nova Scotia
_______________________________________________ Hdf-forum is for HDF software users discussion. [email protected] http://mail.lists.hdfgroup.org/mailman/listinfo/hdf-forum_lists.hdfgroup.org
