Re: Two patches that seem to get gdb working on hppa2.0w-hp-hpux11.00

[Andrew Cagney <ac131313 at cygnus dot com>]

Eric Anderson wrote:
> 
> 1. The HP-UX cc compiler does not seem to like seeing a default: case in
> a switch without something after it; I inserted a break and the
> compiler likes it.

I've posted an up-to-date patch for this. Instead of break, it calls
internal error since the comment claims it should never happen.

> 2. Starting sub-processes in fork-child.c:fork_inferior is done
> without the -f flag, which lets the shell run the startup scripts and
> fork things.  The comments above the execlp seem to indicate that the
> -f should be present, however, it's not there.  I would imagine that
> the problem has been somehow fixed elsewhere, but became un-fixed for
> hppa2.0w-hp-hpux11.00.
> 
> I don't know what the "correct" fix here is; if there is only a
> problem on this particular architecture, you probably only want the
> change on that architecture.
> 
> Without the -f, the code starts the sub-shell, goes into
> startup_inferior, sees a fork, and decrements the count, then sees a
> SIGCHLD, tells the child to resume, but the whole thing wedges there.

I am pretty sure that passing -f to the shell is considered wrong.  If
you do this then other things don't work - such as GDB can no longer
correctly find the program being run.
Others with a better/longer memory are probably in a better position to
give a definitive answer.

	Andrew