On Sat, 22 Oct 2005, Greg Hudson wrote:
> On Sat, 2005-10-22 at 22:20 +0200, Peter N. Lundblad wrote:
> > It might be seen as such, although I think it is a bug that those error
> > messages were just dropped on the floor. Do you object to the change, or
> > do you just think it needs to be documented in the public docstring (the
> > latter, I could agree with in afterthought:).
>
> It seems wrong for a low-level, programmatic API to be interacting with
> stderr.
>
This is a valid point, but I think it is acceptable in this case. A caller
will always need to be prepared for unexpected things on stderr. The
executed application might write to stderr for various reasons, for
example failure to open a shared lib that the program needs.
> An alternative--though a rather more complicated one--would be to open a
> pipe (set FD_CLOEXEC in the child; dunno if you can do that with APR)
> and marshal the error information through it back to the parent, where
> it could be returned to the caller programmatically.
>
I guess this might work as well, it just seems to be overkill. But if you
or others feel strongly about it, I could give it a try.
Thanks,
//Peter
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sun Oct 23 00:06:11 2005