I've opened a bug regarding this http://bugs.freepascal.org/view.php?id=23948 for better tracking this issue. Thanks for all the feedback received here. Vittorio
On Sat, Feb 23, 2013 at 5:50 PM, Vittorio Giovara < vittorio.giov...@gmail.com> wrote: > On 22/feb/2013, at 16:59, Tomas Hajny <xhaj...@hajny.biz> wrote: > > > On Fri, February 22, 2013 16:33, Vittorio Giovara wrote: > > > > > > Hi, > > > >> yes I was talking about linux. I didn't include much information > because I > >> didn't want to confuse people with details. > >> > >> In this branch of Hedgewars > >> https://code.google.com/p/hedgewars/source/list?name=physfslayer we are > >> using physfs library version 2.1.0. > >> However many people out there have version 2.0.3 installed and (of > course) > >> it doesn't contain many of the APIs present in 2.1.0. > >> So when libphysfs is not found (or an old one is detected) we provide a > >> copy of the right physfs and have CMake (our build system) compile and > >> install it. However even if I hardcode the right path of the right > library > >> of the right version, the installed system library takes precedence, and > >> thus if another version of physfs is installed, hedgewars doesn't link > any > >> more. > >> > >> I would refrain from using a different library name; looking at the > >> generated link.res file it seems that the first lib search path is > >> /urs/lib > >> and /lib and the custom one (where the right libphysfs resides) is found > >> only later on. Is there a way to prevent this? I tried to play with the > >> -Xd > >> flag but the result didn't change. > > > > I don't know enough about the library path setting (I leave commenting on > > this to others). However, if you require a specific version number of a > > library (and even distribute that version with your code), explicitly > > linking to that library version seems to be the appropriate solution from > > my point of view. Another option may be linking the respective library > > > > As mentioned, I've been doing only static builds so far. Plus I passed > the linker (with -k) the full path of the library it should link, but > the system one is somehow always dragged in and linked before my > command takes effect. > > Vittorio >
_______________________________________________ fpc-devel maillist - fpc-devel@lists.freepascal.org http://lists.freepascal.org/mailman/listinfo/fpc-devel
