Date: Mon, 21 Aug 2006 17:06:55 -0400 From: Mike Meyer <mwm-keyword-freebsdhackers2.e313df@mired.org> To: des@des.no (Dag-Erling =?iso-8859-1?Q?Sm=F8rgrav?=) Cc: Reko Turja <reko.turja@liukuma.net>, freebsd-hackers@freebsd.org Subject: Re: Aqcuiring full path to running process from outside the kernel Message-ID: <17642.8303.746281.383448@bhuda.mired.org> In-Reply-To: <86sljqnzbz.fsf@xps.des.no> References: <017601c6c486$6477c370$0a0aa8c0@rivendell> <17640.52899.432083.511555@bhuda.mired.org> <86sljqnzbz.fsf@xps.des.no>
next in thread | previous in thread | raw e-mail | index | archive | help
In <86sljqnzbz.fsf@xps.des.no>, Dag-Erling Sm=F8rgrav <des@des.no> type= d: > Mike Meyer <mwm-keyword-freebsdhackers2.e313df@mired.org> writes: > > However, you can also make a best effort. Check argv[0] to see if y= ou > > have an absolute path. If not, do what the shell would do to find t= he > > executable - look for it on $PATH. Once you've found it, call > > realpath(3) to canonicalize the path. > You got it the wrong way around. First pass argv[0] to realpath(3), > fall back to using $PATH only if realpath(3) fails (which it shouldn'= t > unless you've called chdir(2), chroot(2) or jail(2) earlier in the > process, or the executable was moved or removed) No, I got it the right way 'round. If the shell walks the PATH, then calling realpath(3) on argv[0] is the wrong thing to do, as it'll resolve the path relative to the pwd. In particular, since realpath(3) doesn't care if the last component of the path actually exists or not, any argv[0] that doesn't have a '/' in it will be mapped to that name in the current directory. That's assuming that the shell doesn't pass an argv[0] that's the full path to the executable; if the shell does that, then my version still works. Since there are shells that don't do that, you gotta walk the PATH before calling realpath(3) if the shell walked the PATH. Testing for absolute paths is the wrong way test for "did the shell walk PATH", though. You should test for any occurrence of '/' instead, but being absolutely right about it may involve changing the test to match the shell being used. =09<mike --=20 Mike Meyer <mwm@mired.org>=09=09http://www.mired.org/consulting.html Independent Network/Unix/Perforce consultant, email for more informatio= n.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?17642.8303.746281.383448>