[ROOT] CINT: bug -- too many d'tors?

From: Christoph Bugel (chris@uxsrvc.tti-telecom.com)
Date: Sun May 06 2001 - 22:26:44 MEST


Dear Masaharu,

I am using cint for a quite some time now (with pleasure :-), and I
think I found a little bug.  Maybe it is a known limitation?  Anyway, I
think there is a problem that a destructor of a temporary object is
called, even though the constructor was never called. This happens in a
statement with the logical && operator: when the left side is false, the
right side will not be evaluated (according to C/C++ rules). When using
( 1 && ... ) in the attached example, the right side produces a
temporary object, which is later destructed. but when I use ( 0 && ...)
the right side does not produce a temporary object, but the destructor
seems to be called anyway.

I can reproduced the problem with an older version 5.14.50, but also
with 5.14.79 and 5.14.85.  Attached is the code that produces the
problem.  Below is the output with 5.14.85/Linux. The output differs
sometimes, but in all cases the number of d'tors is too high.

cint> { f(); }
foo-ctor: 7
foo-dtor: 7
foo-dtor: 1077300136
foo-dtor: 1077300136
Segmentation fault


Thanks!
Christoph





This archive was generated by hypermail 2b29 : Tue Jan 01 2002 - 17:50:44 MET