All of lore.kernel.org
 help / color / mirror / Atom feed
From: Markus Armbruster <armbru@redhat.com>
To: Peter Maydell <peter.maydell@linaro.org>
Cc: qemu-devel@nongnu.org, patches@linaro.org
Subject: Re: [Qemu-devel] [PATCH 00/12] qemu-common.h/osdep.h refactoring
Date: Thu, 23 Jul 2015 17:44:50 +0200	[thread overview]
Message-ID: <87a8umanf1.fsf@blackfin.pond.sub.org> (raw)
In-Reply-To: <1437583451-9677-1-git-send-email-peter.maydell@linaro.org> (Peter Maydell's message of "Wed, 22 Jul 2015 17:43:59 +0100")

Peter Maydell <peter.maydell@linaro.org> writes:

> This series makes a start at cleaning up some of our headers
> to avoid the common problem of header files including qemu-common.h
> (which then in turn can lead to awkward circular includes).

Yes, please!

> One common cause of this is that we don't have any header
> which will include the basic things most header files require,
> except for qemu-common.h. This series fixes that by making
> 'osdep.h' that "common basic stuff" header. The idea is that:
>  * osdep.h can be included from anywhere, since it doesn't
>    include any other QEMU headers itself except a few very
>    restricted special purpose ones (config-host.h, compiler.h, etc)
>  * osdep.h provides:
>    + things everybody needs, like NULL, int32_t, container_of,
>      the CONFIG_* defines, etc
>    + things that will cause subtle problems if they're not
>      present everywhere (eg directly using system headers and
>      not getting the portability fixups will result in something
>      that builds on most but not all hosts)

I'd file the CONFIG_FOO under "things that may cause subtle problems if
they're not present everywhere" rather than "things everybody needs",
because

1. loads of code doesn't need any, and

2. code that does need it is prone to compile fine but misbehave when
the header providing its CONFIG_FOOs isn't included.

> So most places can just include osdep.h, not the full qemu-common.h.
>
> (I initially thought about defining a new header for this
> purpose, "qemu/basics.h" or some such, but in fact osdep.h
> was already very close to what I wanted so it didn't really
> seem like it was very useful to switch everything over.)
>
> The series has some minor cleanups, some shuffling around of
> things between qemu-common.h, osdep.h and compiler.h, and
> a couple of patches at the end that use osdep.h to allow
> dropping a qemu-common.h include from various header files,
> as a demonstration.
>
> The real question here I guess is whether people like the
> direction I'm trying to go with this. If so, we can further
> reduce the number of qemu-common.h includes without too much
> difficulty with further patches.

For what it's worth, here's how Autoconf wants things done: every .c
file includes config.h *first*, and nothing else includes it.  Easy to
check mechanically, safely avoids the subtle miscompiles mentioned
above.

config.h normally contains just #defines computed by configure, but you
can add whatever else you like, with AH_TOP() or AH_BOTTOM().

It normally is the "things that may cause subtle problems if they're not
present everywhere", but not "things everybody needs (but doesn't want
to include again and again)".  However, nothing stops you from adding
the latter to config.h as well.

My point is:

* Your envisaged role of osdep.h is a common, proven technique.

* Consider requiring every .c to include it before anything else, with
  no other includes allowed.

[...]

  parent reply	other threads:[~2015-07-23 16:23 UTC|newest]

Thread overview: 30+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2015-07-22 16:43 [Qemu-devel] [PATCH 00/12] qemu-common.h/osdep.h refactoring Peter Maydell
2015-07-22 16:44 ` [Qemu-devel] [PATCH 01/12] qapi/qmp-event.c: Don't manually include os-win32.h/os-posix.h Peter Maydell
2015-07-31 12:46   ` Daniel P. Berrange
2015-07-22 16:44 ` [Qemu-devel] [PATCH 02/12] osdep.h: Remove qemu_printf Peter Maydell
2015-07-31 12:46   ` Daniel P. Berrange
2015-07-22 16:44 ` [Qemu-devel] [PATCH 03/12] osdep.h: Move some compiler-specific things to compiler.h Peter Maydell
2015-07-31 12:47   ` Daniel P. Berrange
2015-07-22 16:44 ` [Qemu-devel] [PATCH 04/12] compiler.h: Use glue() in QEMU_BUILD_BUG_ON define Peter Maydell
2015-07-31 12:48   ` Daniel P. Berrange
2015-07-22 16:44 ` [Qemu-devel] [PATCH 05/12] qemu-common.h: Move Win32 fixups into os-win32.h Peter Maydell
2015-07-31 12:49   ` Daniel P. Berrange
2015-07-22 16:44 ` [Qemu-devel] [PATCH 06/12] osdep.h: Move some OS header includes and fixups from qemu-common.h Peter Maydell
2015-07-31 12:50   ` Daniel P. Berrange
2015-07-22 16:44 ` [Qemu-devel] [PATCH 07/12] osdep.h: Add header comment Peter Maydell
2015-07-31 17:31   ` Daniel P. Berrange
2015-07-22 16:44 ` [Qemu-devel] [PATCH 08/12] qemu-common.h: Move muldiv64() to host-utils.h Peter Maydell
2015-07-31 17:32   ` Daniel P. Berrange
2015-07-22 16:44 ` [Qemu-devel] [PATCH 09/12] apic_internal.h: Include cpu.h directly Peter Maydell
2015-07-31 17:33   ` Daniel P. Berrange
2015-07-22 16:44 ` [Qemu-devel] [PATCH 10/12] timer.h: Don't include qemu-common.h Peter Maydell
2015-07-31 17:36   ` Daniel P. Berrange
2015-07-22 16:44 ` [Qemu-devel] [PATCH 11/12] event_notifier.h: " Peter Maydell
2015-07-22 16:44 ` [Qemu-devel] [PATCH 12/12] throttle.h: " Peter Maydell
2015-07-22 20:09 ` [Qemu-devel] [PATCH 00/12] qemu-common.h/osdep.h refactoring Paolo Bonzini
2015-07-23 15:44 ` Markus Armbruster [this message]
2015-07-31 12:54 ` Daniel P. Berrange
2015-07-31 12:57   ` Peter Maydell
2015-07-31 17:30     ` Daniel P. Berrange
2015-08-13 17:37 ` Peter Maydell
2015-08-19 18:18   ` Peter Maydell

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=87a8umanf1.fsf@blackfin.pond.sub.org \
    --to=armbru@redhat.com \
    --cc=patches@linaro.org \
    --cc=peter.maydell@linaro.org \
    --cc=qemu-devel@nongnu.org \
    /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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.