swat ! String pointers & VarProbes

[Sven Thommesen]

Since I burdened the list with my recent core dump troubles,
I guess I'd better report that the problem has been solved.
And <blush, blush> it was all my fault! 

Despite the way some C books make it look like character arrays 
and character pointers are 'the same', they're not.  If you define

char * X;

and then use sprintf(X,...) or strcpy(X,...) without first initalizing X to
point to a valid string, the results are ... a core dump, and one that gdb
is of 
little help with, to boot.

And when K&R (2nd ed, p. 194) say that "the behavior of a program that
attempts to alter a string literal is undefined", they mean it. Undefined,
as in another core dump!

(That is, defining 'char * X = "SomeString" ' and then using sprintf or
strcpy with X as a target.)

My thanks to Glen for handholding.

We did, trying to track down this elusive bug, discover one interesting
thing about the VarProbe machinery:  it works the opposite way of sprintf()
and strcpy() ! If you allocate an instance variable 'char X[80]' in your
program and try to use ObjectLoader to fill that variable from a disk file,
you get an error message. No can do. But if you define 'char * X' and leave
it un-initialized, the probe code will load your string into memory
somewhere and make X point to it. Go figure.


Hope this confessional was of marginal use to someone.

Now on to the next bug, wherever it may lurk!

Sven.