On Fri, 11 Sep 2015 17:12:50 +0300 Valentin Davydov <sqlite-user at soi.spb.ru> wrote:
> On Thu, Sep 10, 2015 at 01:30:37PM +0200, Eduardo Morras wrote: > > > > Use gmake to compile. > > It didn't work either. Finally I've just installed some brand new > linux on a nearby virtual machine, made there make -f > Makefile.linux-gcc and thoroughly repeated it's output line-by-line > on my FreeBSD while replacing gcc by cc. Eventually it compiled well > (with almost the same set of warnings) and the result seems to work > (at least within my own coverage). > > Thanks to the high general coding style, there were only two > idiosyncrasies to be corrected: lack of the system-wide malloc.h > (which is replaced by unistd.h in FreeBSD) and yet another lack of > the fdatasync() syscall, which is already carefully screened from the > sources by a special symbol called __ANDROID__ ;-) No, malloc.h is on libc stdlib.h Fdatasync() syscall is linuxism and don't exist in FreeBSD, use fsync. Note that FreeBSD filesystem UFS2 softupdates takes care of separate metadata and data parts of a file write/update/delete. > By the way, clang kindly revealed a couple of forgotten "unsigned" in > the sources, which could (and did in the past) produce some nice > bugs. Attention to developers. > > > Note that there isn't a port, > > I know. Despite the code being frozen for almost a full year, nobody > wants to take a trouble of maintaining FreeBSD port (perhaps me too). It's experimental code, I played with it before, but don't use on any project. A port has no sense for now, I think. > Valentin Davydov. --- --- Eduardo Morras <emorrasg at yahoo.es>
