Date: Tue, 21 Nov 2006 12:50:53 -0600 (CST) From: "Sean C. Farley" <sean-freebsd@farley.org> To: Stefan Esser <se@freebsd.org> Cc: freebsd-current@freebsd.org Subject: Re: Fix for memory leak in setenv/unsetenv Message-ID: <20061121124316.Q4388@baba.farley.org> In-Reply-To: <20061019130354.D92319@thor.farley.org> References: <20061006200320.T1063@baba.farley.org> <200610111427.42195.jhb@freebsd.org> <20061018211005.L1466@baba.farley.org> <200610191032.29232.jhb@freebsd.org> <20061019112601.J91957@thor.farley.org> <4537B53A.3080208@FreeBSD.org> <20061019130354.D92319@thor.farley.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Thu, 19 Oct 2006, Sean C. Farley wrote: > On Thu, 19 Oct 2006, Stefan Esser wrote: > > I've got to admit, that I have not looked your patch, but the only > > drawback seems to be that the last instance of a variable in the > > environment space has to be located in getenv() (maximizing the > > search time ...). > > Actually, the first active variable found is returned by getenv() even > if another would be found later. This does make me think that if > changed the way the environment variable array was built to only > contain the first instance of each variable instead of all instances > then a search by getenv() from the end of the array backwards would be > faster. A cheap alternative is to create the array in reverse. > > > Always using the last allocated (largest) slot for storage of new > > values of environment variables would result in nearly reasonable > > behavior. A cached pointer does either point to the value of the > > variable at the time of the getenv(), or to the last value assigned > > to the environment variable that does not exceed the allocated size. > > I will look at changing it. This is a resend from two weeks ago with an edit about using a constructor. I changed it to create the array in reverse and to search it from the end. I played with crt1.c to have it initialize the new environment and clean it up via atexit(). The feeling I get is that this may be disliked since malloc_init() is not called here. Would there be a reason not to have things such as memory allocation and an environment (such as I have been doing) initialized and/or cleaned up here? How about using a constructor (__attribute__((constructor)))? Would there be anything else I should code to have this be a viable replacement to the current implementation of the setenv-family of functions? Sean -- sean-freebsd@farley.org
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20061121124316.Q4388>