Re: bug in rootcint

From: Philippe Canal <pcanal_at_fnal.gov>
Date: Thu, 4 Sep 2008 11:20:21 -0500


Hi Robert,

Did you try running rootcint under valgrind? Where/How can I reproduce this problem?

Thanks,
Philippe

Robert Hatcher wrote:
> Hi all,
> For about the last week I've been having problems with the SVN trunk
> of rootcint. The error looks like:
>
> $ rootcint -f /afs/path/tmp/UtilCint.cc -c \
> -I. -I/afs/path/tmp -I/afs/path/tmp -I/afs/path/rel/include \
> -I/afs/path/rel/include -DUNIX -DLINUX -D__UNIX__ -D__LINUX__ \
> -DDATAREP_LITTLE_IEEE -DDATAREP_LITTLE_ENDIAN \
> -DDEFECT_NO_IOSTREAM_NAMESPACES -DDEFECT_NO_JZEXT -DDEFECT_NO_INTHEX \
> -DDEFECT_NO_INTHOLLERITH -DDEFECT_NO_READONLY -DDEFECT_NO_DIRECT_FIXED \
> -DDEFECT_NO_STRUCTURE -DUSE_ROOT -I/afs/path/root/trunk-opt/include \
> LoadMinosPDG.h MsgUtil.h UtilBlockedRunningAverage.h \
> UtilBlockedRunningAverageVector.h UtilHepevt.h UtilIstHEP.h \
> UtilMCFlag.h UtilMath.h UtilPDG.h UtilRunningAverage.h \
> UtilRunningAverageVector.h UtilStream.h UtilString.h LinkDef.h
> Error: link requested for unknown class UtilPDG LinkDef.h:28:
> Warning: Error occurred during reading source files
> Warning: Error occurred during dictionary source generation
> !!!Removing /afs/path/tmp/UtilCint.cc /afs/path/tmp/UtilCint.h !!!
> Error: rootcint: error loading headers...
>
> This is on Linux SLF3.0.5 w/ gcc 3.4.3. I can't come up with a
> consistent test of this problem. Relevant notes:
> * Our code (in this pkg) hasn't changed, only root versions
> which update nightly. This problem was first observed after the
> clean rebuild on Sunday night (other nights are incremental builds).
> * This error occurs _only_ for the version of root built w/
> optimization turned on. Our non-opt build of root against
> this same instance of the same package is fine.
> * One of the packages (referencing these same classes) that
> was giving problems cleared up after root was re-built yesterday
> following a "distclean" (as opposed to the nightly incremental build)
> * The problematic files have namespaces, as opposed to classes, with
> enums, typedefs and functions.
>
> The fact that the problem moves around depending on library loading
> order and for the optimized version suggests that the error lies with
> an off-by-one error or a missing \0 in string handling that depends on
> memory layout.
>
> -robert
>
> Robert W. Hatcher | rhatcher_at_fnal.gov
> <mailto:rhatcher_at_fnal.gov> 630-840-3102
> FNAL CD/REX (MINOS) | MS 220, PO Box 500, Batavia IL 60510
>
>
>
>
Received on Thu Sep 04 2008 - 18:20:31 CEST

This archive was generated by hypermail 2.2.0 : Fri Sep 05 2008 - 17:50:02 CEST