Skip to contentSkip to conversation

Kill snd_assert()

DISCUSSION Takashi Iwai 2 participants 4 messages Aug 11, 2008

lore
Share & tools
lore
#0 lore raw

Hi,

as some people already mentioned, we have a pretty ugly macro, snd_assert(). Actually, this macro is sometimes useful, but the implementation including the code flow in arguments is really bad.

I tried to kill this, either the following way:

- replace

	snd_assert(cond, flow...);
   with
	if (snd_BUG_ON(!cond))
		flow...;

- use WARN_ON() if the check is needed essentially regardless of CONFIG_SND_DBEBUG value

- remove snd_assert() for obviously unneeded code-paths

The new snd_BUG_ON() macro takes the condition, and works like WARN_ON() if CONFIG_SND_DEBUG=y: prints the warning with stack trace, and returns the evaluated value. When CONFIG_SND_DEBUG=n, the macro returns always zero, so that the compiler will optimize out the unreached flows.

Since snd_BUG_ON() takes the condition of a bug, it's a negative condition against snd_assert(). It follows BUG_ON() and WARN_ON(), and easier to understand, IMO.

The patches are found on topic/remove-assert branch of my git tree:

git://git.kernel.org/pub/scm/linux/kernel/tiwai/sound-2.6.git

For gitweb,
http://git.kernel.org/?p=linux/kernel/git/tiwai/sound-2.6.git;a=shortlog;h=topic/remove-assert

I haven't merged this topic yet. If you have a big objection or a better implementation idea, please let me know.

Takashi

#1 lore raw
Quoted 6 lines · Takashi Iwai

Hi,

as some people already mentioned, we have a pretty ugly macro, snd_assert(). Actually, this macro is sometimes useful, but the implementation including the code flow in arguments is really bad.

Quoted 7 lines

I tried to kill this, either the following way:

- replace

	snd_assert(cond, flow...);
   with
	if (snd_BUG_ON(!cond))
		flow...;

For my eyes, many uses of snd_assert with direct 'return VALUE' are more "sexy", easy understandable and smaller. But, as mentioned, when goto is used to change the code flow, we end up with a compiler warning. It should be eliminated.

I have no strong objections to accept this change as you proposed. Thank you for your work.

Jaroslav

-----
Jaroslav Kysela <[email protected]>
Linux Kernel Sound Maintainer
ALSA Project, Red Hat, Inc.

#2 lore raw

At Mon, 11 Aug 2008 10:52:23 +0200 (CEST),

Quoted 11 lines · Jaroslav Kysela
Quoted 8 lines · Takashi Iwai

Hi,

as some people already mentioned, we have a pretty ugly macro, snd_assert(). Actually, this macro is sometimes useful, but the implementation including the code flow in arguments is really bad.

I tried to kill this, either the following way:

Quoted 8 lines
Quoted 5 lines

- replace

	snd_assert(cond, flow...);
   with
	if (snd_BUG_ON(!cond))
		flow...;

For my eyes, many uses of snd_assert with direct 'return VALUE' are more "sexy", easy understandable and smaller.

There are pros and cons for this, too. One merit of snd_assert() is its shortness, thus easier to add. But, mining the code flow is dangerous especially if you want to add a paired operation like locks or move the assert to another place. With the explicit code flow, you could follow more easily.

Quoted 3 lines

But, as mentioned, when goto is
used to change the code flow, we end up with a compiler warning. It should be eliminated.

Right.

Quoted 2 lines

I have no strong objections to accept this change as you proposed. Thank you for your work.

OK, good to hear.

thanks,

Takashi

#3 lore raw
In reply to Takashi Iwai

At Mon, 11 Aug 2008 10:41:37 +0200,

Quoted 26 lines · I wrote:

Hi,

as some people already mentioned, we have a pretty ugly macro, snd_assert(). Actually, this macro is sometimes useful, but the implementation including the code flow in arguments is really bad.

I tried to kill this, either the following way:

- replace

	snd_assert(cond, flow...);
   with
	if (snd_BUG_ON(!cond))
		flow...;

- use WARN_ON() if the check is needed essentially regardless of CONFIG_SND_DBEBUG value

- remove snd_assert() for obviously unneeded code-paths

The new snd_BUG_ON() macro takes the condition, and works like WARN_ON() if CONFIG_SND_DEBUG=y: prints the warning with stack trace, and returns the evaluated value. When CONFIG_SND_DEBUG=n, the macro returns always zero, so that the compiler will optimize out the unreached flows.

Quoted 8 lines

Since snd_BUG_ON() takes the condition of a bug, it's a negative condition against snd_assert(). It follows BUG_ON() and WARN_ON(), and easier to understand, IMO.

The patches are found on topic/remove-assert branch of my git tree:

   git://git.kernel.org/pub/scm/linux/kernel/tiwai/sound-2.6.git

Oops, the correct URL is:

    git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound-2.6.git

Also, the corresponding alsa-driver build tree is found on topic/remove-assert branch of alsa-driver-build.git below:
git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/alsa-driver-build.git

http://git.kernel.org/?p=linux/kernel/git/tiwai/alsa-driver-build.git;a=shortlog;h=topic/remove-assert

Takashi