Re: [PATCH 2/6] Introduce throw_ptrace_error

[Mark Kettenis <mark.kettenis@xs4all.nl>]

> Date: Fri, 06 Mar 2015 21:40:03 +0000
> From: Pedro Alves <palves@redhat.com>
> 
> On 03/06/2015 09:03 PM, Mark Kettenis wrote:
> >> From: Pedro Alves <palves@redhat.com>
> >> Date: Fri,  6 Mar 2015 19:58:02 +0000
> >>
> >> This adds a new function that is meant to be called instead of
> >> perror_with_name whenever we get an error out of ptrace.  The
> >> idea is to convert some errno errors to different GDB exceptions in
> >> a following patch.
> > 
> > What is ptrace-specific about throwing an error?  This really feels
> > like the wrong direction to me.
> 
> Not exactly sure what you mean.  Throwing an error is
> of course not ptrace-specific.  What is ptrace-specific is the
> interpretation of errno.  The end result of the series is that
> an ESRCH as a result of a ptrace error ends up throwing
> a THREAD_NOT_FOUND_ERROR error instead of a GENERIC_ERROR.  Then
> callers up the chain can actually distinguish the errors.
> 
> Please take a look at patch 3.  It should make things clearer.

I think your interpretation of ESRCH is too Linux-centric.  You're
once again duct-taping around the Linux kernel's whoefully
insufficient threads debugging capabilities.  It really should not be
possible for a thread to just disappear without the debugger being
notified.  Do I sound like a broken record?

I think at this point the right approach is to make
linux_resume_one_lwp() call ptrace() directly instead of calling down
into the inf_ptrace_resume().  That way you can simply check errno in
the place where it matters.