From: Simon Marchi <simark@simark.ca>
To: Tom Tromey <tom@tromey.com>
Cc: gdb-patches@sourceware.org
Subject: Re: [PATCH 3/4] Change thread_to_thread_object to return a new reference
Date: Mon, 17 Sep 2018 00:52:00 -0000 [thread overview]
Message-ID: <03a3606359da188a8100201c7dfae823@simark.ca> (raw)
In-Reply-To: <875zz5pmg8.fsf@tromey.com>
On 2018-09-16 10:05, Tom Tromey wrote:
>>>>>> "Simon" == Simon Marchi <simark@simark.ca> writes:
>
>>> + if (result == NULL)
>>> + result = gdbpy_ref<>::new_reference (Py_None);
>
> Simon> I would suggest using Py_RETURN_NONE, which we already use at
> many places.
>
> Simon> Instead of setting that result variable, is there something
> preventing you
> Simon> from returning directly above, like this?
>
> Simon> if (thread_info != NULL)
> Simon> return thread_to_thread_object (thread_info).release ();
>
> Simon> The bottom line could just be unconditionally Py_RETURN_NONE.
>
> I made this change.
>
> In the past you couldn't return from inside a try/catch but I think it
> is ok now.
>
>
> I couldn't use Py_RETURN_NONE in py_get_event_thread because it returns
> a gdbpy_ref<> -- and I think the latter is more important.
Oh right. While working on this code I though adding a
GDB_Py_REF_RETURN_NONE macro, which would be like Py_RETURN_NONE but
returning a gdbpy_ref. Not sure it's really that useful, since what you
wrote is explicit and clear.
> I re-read all of this and I think thread_to_thread_object has a latent
> bug. It can return NULL early due to an error:
>
> gdbpy_ref<inferior_object> inf_obj (inferior_to_inferior_object
> (thr->inf));
> if (inf_obj == NULL)
> return NULL;
>
> or it can return NULL at the end meaning "thread not found".
I am not sure how these two mode of failure differ. Don't we expect the
passed thread_info object to always be valid, and therefore both cases
returning NULL would be an internal GDB logic error?
> I think it is best to have a single style - returning NULL should also
> set the Python exception.
That makes sense, so the caller can just check for NULL and return NULL
right away (as the snippet above does). We should expect
inferior_to_inferior_object to set the exception on failure.
> It seems to me that the second NULL return should just set an
> exception. It's not a normal case.
Right.
Responding directly to your next message:
> Maybe the gdbpy_ref (1-argument) constructor and release methods could
> assert that the Python exception is set if the underlying pointer is
> NULL. That would not get full checking but maybe it would catch some
> problems. And maybe we should simply use gdbpy_ref in many more places
> in the Python layer -- ideally, reserve raw pointers solely for
> parameters which are borrowed references.
Makes sense.
> Let me know what you think of this.
>
> My apologies again for missing this review and pushing too eagerly.
No problem, LGTM.
Simon
next prev parent reply other threads:[~2018-09-17 0:52 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2018-09-13 5:30 [PATCH 0/4] Disallow the return of borrowed references Tom Tromey
2018-09-13 5:30 ` [PATCH 3/4] Change thread_to_thread_object to return a new reference Tom Tromey
2018-09-16 2:11 ` Simon Marchi
2018-09-16 13:32 ` Tom Tromey
2018-09-16 14:05 ` Tom Tromey
2018-09-16 15:35 ` Tom Tromey
2018-09-17 0:52 ` Simon Marchi [this message]
2018-09-17 5:31 ` Tom Tromey
2018-09-13 5:30 ` [PATCH 4/4] Remove CPYCHECKER_RETURNS_BORROWED_REF Tom Tromey
2018-09-13 5:30 ` [PATCH 2/4] Change objfile_to_objfile_object to return a new reference Tom Tromey
2018-09-16 1:28 ` Simon Marchi
2018-09-16 2:00 ` Simon Marchi
2018-09-13 5:30 ` [PATCH 1/4] Change pspace_to_pspace_object " Tom Tromey
2018-09-16 0:57 ` Simon Marchi
2018-09-16 12:59 ` Tom Tromey
2018-09-16 1:19 ` Simon Marchi
2018-09-16 1:58 ` Simon Marchi
2018-09-16 13:01 ` Tom Tromey
2018-09-16 0:56 ` [PATCH 0/4] Disallow the return of borrowed references Simon Marchi
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=03a3606359da188a8100201c7dfae823@simark.ca \
--to=simark@simark.ca \
--cc=gdb-patches@sourceware.org \
--cc=tom@tromey.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).