[pacman-dev] Bug in libalpm: sizeof(off_t)

[Allan McRae]

On 12/12/13 02:42, Jeremy Heiner wrote:
> On Wed, Dec 11, 2013 at 11:14 AM, Allan McRae <allan at archlinux.org> wrote:
>> How about #5...
> 
> Remember: libalpm is a .so, and sizeof(off_t) is something the app
> writer is free to choose (on many systems). The problem is that the
> choice made by libalpm (by ./configure) gets baked into the .so ABI,
> and if the app writer chooses differently they get SEGVs.

Yes.  That is why I was proposing an approach that had sizeof(off_t)
baked in and a second calculation done at runtime.  It turns out my
"runtime" calculation was easily optimized out...   Fixed version on the
way.

A