Hey Peter, I'm fine with your version of patch and noone else had objections so I believe we can patch it your way.
Did you fix it upstream? I'm still seeing 1.11 as the latest release. I can patch it only for Debian but I believe we should have it fixed for every other distros / ways of distribution. regards Bartek W dniu 3/21/2015 o 12:28 AM, Peter Selinger pisze: > Here's the patch that I am planning to apply upstream. Please comment > if you see anything wrong with it. > > While the general idea is similar to Tomasz's patch, I've solved the > details a bit differently. > > * I prefer to use ssize_t instead of unsigned long long int for memory > manipulations. Since size_t is the type used by malloc, memcpy, etc, > it is big enough to hold the relevant values. The reason I use a > signed rather than unsigned type is that the "dy" field in the > potrace_bitmap_s structure may be positive or negative, depending on > whether the bitmap is stored top-to-bottom or bottom-to-top. Potrace > itself always uses a positive dy, but other applications that link > against the Potrace library may use their own convention. Tomasz's > patch used an unsigned type which would break applications that use > a negative dy. > > The code now checks that the bitmap dimensions are indeed such that > all relevant values fit within ssize_t. A remaining assumption is > that ssize_t is at least as big as int, which I think is guaranteed. > > * I prefer to use calloc instead of safe_malloc. Calloc is appropriate > whenever the memory to be allocated is a number of copies of items > of a given size. Unlike malloc(x*y), calloc(x, y) actually checks > that x*y does not overflow. (I checked the glibc source code for > calloc to be sure that such a check is actually performed). In the > few cases where the argument of malloc is calculated differently > (say as a product of three numbers), I have added an explicit > overflow check. This is safer, in my opinion, than safe_malloc(x*y); > in particular, there is no difference bewteen safe_malloc and > ordinary malloc when size_t = unsigned long long int. > > * I also fixed analogous issues in Mkbitmap and throughout the rest of > the code. > > I'll post an updated upstream package in a day or two unless there's > feedback requiring additional changes. > > Thanks, -- Peter > -- To UNSUBSCRIBE, email to debian-bugs-dist-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org
