------- Comment #2 from skinner at milkyway dot gsfc dot nasa dot gov 2008-01-18 00:49 ------- Subject: Re: gfortran crashes and asks for bug report
burnus at gcc dot gnu dot org wrote: > ------- Comment #1 from burnus at gcc dot gnu dot org 2008-01-17 22:45 > ------- > Thanks for the bug report; however, I fail to reproduce the bug with either of > 4.1.3, 4.2.2 and 4.3.0 on x86-64-linux. > > Your version "4.2.0 (experimental)" is relatively old. (Even 4.1.3 might > contain a bug fix, which was not in yet in 4.2.0.) > > Can you try with a newer version, for instance the one available from > http://gcc.gnu.org/wiki/GFortranBinaries ? > > * * * > > The newest gfortran versions also have the following warning by default: > > weights = rate_err**-2.0 > 1 > Warning: Extension: Unary operator following arithmetic operator (use > parentheses) at (1) > > This is not only invalid Fortran but also not always well defined. What does > the following mean: "a**-b*c" Does mean "a**(-b) * c" or "a**(-b*c)" - > different compilers make different choices here. (In your case the choice is > unambiguous but still not standard conform.) > > > Thanks for the v quick response I have installed gcc version 4.3.0 20071231 (experimental) [trunk revision 131236] (GCC) and the error goes away. I can turn it on and off by switching between the two, though it is not conclusive that the problem has been solved because it appeared to depend in some complex way on how the program was arranged, as if it was an overwriting/ buffer-overflow/memory leak sort of problem. So its disappearance could be a chance of how memory is allocated. Anyway my objective was to keep you informed - I had already found a fudge which allowed me to get the results I needed from the quick and dirty program. Incidentally the new version does indeed warn about **-2.0 . I agree it is bad practice but without looking carefully at the rules I am not sure that it is ungrammatical. -2.0 is after all a constant, so the expression is of the form [variable] [operator] [constant] Gerry -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34837
