This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: [PATCH 1/2] Warn when accessing binaries over RSP
- From: Gary Benson <gbenson at redhat dot com>
- To: Andrew Burgess <andrew dot burgess at embecosm dot com>
- Cc: gdb-patches at sourceware dot org, Sandra Loosemore <sandra at codesourcery dot com>, Doug Evans <dje at google dot com>, Pedro Alves <palves at redhat dot com>, Jan Kratochvil <jan dot kratochvil at redhat dot com>, André Pönitz <apoenitz at t-online dot de>, Paul_Koning at Dell dot com
- Date: Tue, 11 Aug 2015 15:04:40 +0100
- Subject: Re: [PATCH 1/2] Warn when accessing binaries over RSP
- Authentication-results: sourceware.org; auth=none
- References: <1438788496-32246-1-git-send-email-gbenson at redhat dot com> <1438788496-32246-2-git-send-email-gbenson at redhat dot com> <20150811115538 dot GA7352 at embecosm dot com>
Andrew Burgess wrote:
> * Gary Benson <gbenson@redhat.com> [2015-08-05 16:28:15 +0100]:
> >
> > diff --git a/gdb/gdb_bfd.c b/gdb/gdb_bfd.c
> > index 1781d80..b511777 100644
> > --- a/gdb/gdb_bfd.c
> > +++ b/gdb/gdb_bfd.c
> > @@ -219,13 +219,38 @@ gdb_bfd_iovec_fileio_open (struct bfd *abfd, void *inferior)
> > const char *filename = bfd_get_filename (abfd);
> > int fd, target_errno;
> > int *stream;
> > + struct target_ops *ops = find_target_at (process_stratum);
> >
> > gdb_assert (is_target_filename (filename));
> > + filename += strlen (TARGET_SYSROOT_PREFIX);
> > +
> > + /* GDB provides no indicator of progress during file operations, and
> > + can appear to have locked up during slow remote transfers, so we
> > + inform the user what is happening and suggest a way out. It's
> > + unpleasant that we need to detect remote targets this way (rather
> > + than putting the warnings in remote_hostio_open), but it's not
> > + possible for remote_hostio_open to differentiate between
> > + accessing inferior binaries (which can be bypassed) and accessing
> > + things like /proc/ (which is unavoidable). */
> > + if (strcmp (ops->to_shortname, "remote") == 0
> > + || strcmp (ops->to_shortname, "extended-remote") == 0)
> > + {
> > + static int warning_issued = 0;
> > +
> > + printf_unfiltered (_("Reading %s from remote target\n"),
> > + filename);
> > +
> > + if (!warning_issued)
> > + {
> > + warning (_("File transfers from remote targets can be slow.\n"
> > + "Use \"set sysroot\" to access files locally"
> > + " instead."));
> > + warning_issued = 1;
> > + }
> > + }
>
> Altering the behaviour based on to_shortname feels like breaking
> this nice target OO model we have.
Yeah... :-/
> Could the warning not be moved down into target_fileio_open instead?
Not so much target_fileio_open as remote_hostio_open; only remote
targets need the warning. And originally I thought no, the warning
couldn't go there, because target_fileio_open/remote_hostio_open is
used for various internal things such as /proc/ file reads on Linux
that the user shouldn't see.
...however...
remote_hostio_open *can* differentiate between reading inferior
binaries and reading internal stuff because the internal stuff is
accessed with the INF argument NULL and binaries are accessed with
a non-NULL INF.
So I can do that, if it doesn't seem too hacky.
> Or if that's really not an appropriate place should we add a new
> target method?
I considered that but couldn't think of a good name :-)
target_fileio_warn_if_slow ??
I can do that too.
Cheers,
Gary
--
http://gbenson.net/