Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 11 Oct 2005 19:34:55 -0400
From:      Mikhail Teterin <mi+mx@aldan.algebra.com>
To:        Daniel Eischen <deischen@freebsd.org>
Cc:        threads@freebsd.org, re@freebsd.org
Subject:   Re: debug defines in libc_r on 6.0
Message-ID:  <200510111934.56300.mi%2Bmx@aldan.algebra.com>
In-Reply-To: <Pine.GSO.4.43.0510111856210.7953-100000@sea.ntplx.net>
References:  <Pine.GSO.4.43.0510111856210.7953-100000@sea.ntplx.net>

next in thread | previous in thread | raw e-mail | index | archive | help
צ×ÔÏÒÏË 11 ÖÏ×ÔÅÎØ 2005 19:00, Daniel Eischen ÷É ÎÁÐÉÓÁÌÉ:
> No, it's only a couple of instructions for a very few operations.
> I've already replied to you about these before. šI don't care
> whether re@ turns them off or not, but go read my prior email
> on this topic. šI should eliminate the #defines and always do
> the checks, so you wouldn't notice :(

We make such checks (however inexpensive) conditional on various DEBUG* and 
INVARIANTS* flags in the kernel and userspace should be no different. The 
whole idea of assert() is to provide a way to add debugging checks, which -- 
however cheap they are -- can be easily removed from a _release_ version of 
one's software.

I remember your e-mail and continue to disagree with it -- the debugging 
information and checks are, understandably, precious to you, but if we lose, 
say, a MySQL benchmark by even a small percentage, the advocacy loss will be 
immense...

IMHO, the flags should be removed in RELENG_6, or, at least, in RELENG_6_0.

Respectfully,

	-mi



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?200510111934.56300.mi%2Bmx>