Date: Tue, 23 Jun 2009 02:27:15 +0300 From: Andriy Gapon <avg@icyb.net.ua> To: Adrian Chadd <adrian@freebsd.org> Cc: Roman Divacky <rdivacky@freebsd.org>, current@freebsd.org, Steve Kargl <sgk@troutmask.apl.washington.edu> Subject: Re: [PATCH]: if (cond); foo() in firewire Message-ID: <4A401353.6070703@icyb.net.ua> In-Reply-To: <d763ac660906221610j5aa8f085uc78be271d98f1c11@mail.gmail.com> References: <20090621082022.GA88526@freebsd.org> <d763ac660906212104w4bf066eatf5529779e603bd0e@mail.gmail.com> <20090622045428.GA18123@troutmask.apl.washington.edu> <4A3F6917.7040806@icyb.net.ua> <d763ac660906221610j5aa8f085uc78be271d98f1c11@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
on 23/06/2009 02:10 Adrian Chadd said the following: > 2009/6/22 Andriy Gapon <avg@icyb.net.ua>: >> You confuse me. It is a "vanilla userland transfer", but so? >> Current code always goes to "out" label regardless if uimove succeeded or not. >> I think the idea was to go "out" only if uimove failed and execute some code >> between if and out-label otherwise. > > Because now you have a code path being run which hasn't been run for > quite a while. > > I'm just saying be careful, and don't assume that "clang found a bug". > It found a bad code construct. Changing that bit of code changes the > flow of execution and may change things unexpectedly in later code. > It's the same with any bug - this "found by clang" bug should be > looked at by someone who knows the firewire code and they haven't > replied to this thread. :) I must agree with you, no other choice. But my thinking is this: let's fix the obvious typo (I am sure-sure that this is what it is) and then let any "real" bugs (if any) bite firewire guys the hard way. I.e. if the choice is between: 1) fix the typo now and potentially provoke dormant bugs; 2) indefinitely wait and don't fix the typo until anybody comes forward and declares that there are no dormant bugs in the vicinity; then I'd choose #1. > I'm glad clang has this lexical analysis magic. Shouldn't there be > some kind of weird, magical, standalone "lint" program to do this kind > of lexical checking for us? :) I guess there should be one. But as simple as C language standard is :-) it seems that even with a magnitude of tools we are bound to only approximate the perfection. -- Andriy Gapon
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4A401353.6070703>