Re: __attribute__ ((cleanup) patch
- From: Colin Walters <walters verbum org>
- To: Pavel Simerda <psimerda redhat com>
- Cc: networkmanager-list gnome org
- Subject: Re: __attribute__ ((cleanup) patch
- Date: Thu, 18 Oct 2012 11:51:09 -0400
On Thu, 2012-10-18 at 11:19 -0400, Pavel Simerda wrote:
> Then why it wasn't good enough for Glib but is good enough for NetworkManager? Why
See this thread:
https://mail.gnome.org/archives/gtk-devel-list/2012-April/msg00003.html
Basically, because the GTK+ stack needs to compile with MSVC.
Unfortunately. And Microsoft said they'll never implement anything
beyond C89, not even C99.
> I don't see a link to elaborate explanation and documentation?
I'll add some more docs to libgsystem. But the most important thing
to read first is
http://gcc.gnu.org/onlinedocs/gcc/Variable-Attributes.html
> Success stories.
Like I said, systemd and upstart both use it, and I completely rely on
it inside ostree (http://git.gnome.org/browse/ostree) which is 15,000
lines of GObject/C. A tenth the size of NM, true, but also not a
trivial project.
We're using it in another currently Red Hat internal project. I think
Ray is amenable to doing it for GDM.
Ultimately though, the only way the library is going to get better is
if people use it and help out.
> More
> information?
I did blog a bit ago about it too:
http://blog.verbum.org/2012/05/09/__attribute__-cleanup-or-how-i-came-to-love-c-again/
> That's exactly the problem I have with such quick inclusions without a proper discussion.
Fair enough.
> I'm not yet even convinced about this because of total lack of documentation to be
> found right away.
Ok, I'll add some more.
> And I'm not convinced that switching to a non-standard programming
> language is the price to pay.
It's kind of a leap to call this a "non-standard programming language".
In reality of course, GLib (which NM relies on) is a *heavy* user of GCC
extensions. Things like atomic intrinstics and statement expressions.
This is larger, too, but it's really just C with one feature from C++.
> So, I haven't changed my mind. I propose the following path:
>
> 1) Revert
I don't oppose that.
> 2) Document including the reasons why this cannot be added to Glib but should be
> added to NetworkManager
Already done here.
> 3) Discuss
In progress =)
[
Date Prev][
Date Next] [
Thread Prev][
Thread Next]
[
Thread Index]
[
Date Index]
[
Author Index]