Date: Fri, 13 May 2016 09:42:30 -0700 (PDT) From: Don Lewis <truckman@FreeBSD.org> To: brde@optusnet.com.au Cc: src-committers@freebsd.org, svn-src-all@freebsd.org, svn-src-head@freebsd.org Subject: Re: svn commit: r299588 - head/usr.sbin/binmiscctl Message-ID: <201605131642.u4DGgUt5060836@gw.catspoiler.org> In-Reply-To: <20160514013315.K1048@besplex.bde.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On 14 May, Bruce Evans wrote: > On Fri, 13 May 2016, Don Lewis wrote: > >> Log: >> Revert r299584: >> Mark usage() as __dead2 so that Coverity doesn't think that execution >> continues after the call and uses a negative array subscript. >> >> Requested by: bde > > Thanks. > > What was the problem that confused Coverity? It sees that demux_cmd() can return -1. It takes the true branch of "if (cmd == -1)" and calls usage. It then falls through and thinks that cmds[] is getting indexed with a negative value on line 423. Taking a closer look at the comments that Coverity added makes me wonder if changing the test to "if (cmd < 0)" might unconfuse it. > This reminds me that even compilers can see that usage() doesn't return, > by looking ahead and even inlining usage(). This breaks debugging and > profiling. clang doesn't support the -funit-at-a-time and > -fno-inline-functions-called-once needed to reduce its excessive inlining.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201605131642.u4DGgUt5060836>