Re: [RFC] Re: Profiling gdb?

[Andrew Cagney <ac131313@redhat.com>]

> 
> _mcleanup() is no more or less published than monstartup(), as far as I
> can tell; I'll submit a glibc patch to document these when I get around
> to it.  I know both FreeBSD and GNU/Linux have these, because that's
> how Emacs controls profiling.

Good old emacs - can always be relied on as the keeper of portability 
knowledge.

FYI,

They are all based on the same berkley code.  Looking at the Free/Net 
BSD man pages they document.

      moncontrol(int mode);

      monstartup(u_long *lowpc, u_long *highpc);

but not _mcleanup() (hence my comment about interfaces).  One thing, 
while the doco indicates pointers, the actual code matches your 
declaration.  So gess we stick with that.

http://www.opengroup.org doesn't mention any of this.

glibc recently modified its internal structures so that are incompatible 
with the other implementations.

FreeBSD also modified the internal structures but in a `forward 
compatible' way.

NetBSD hasn't touched the code in yonks (other than to fix portability 
problems).

> Hmm, make that less documented, since FreeBSD has a man page for
> monstartup.  They're both provided, however.
> 
> I am indeed sure of monstartup()'s function signature - well, at least
> for GNU/Linux; I don't have access to a FreeBSD system but according to
> the source for libc in FreeBSD 5.0 it'll work there too.  Although I
> think it may leak memory in multiple runs, the way I use it.  It's
> declared in a non-system-specific file as taking longs (unsigned, oops,
> will fix).

Andrew