On Sat, 02 Aug 2003 00:39:58 +0200 CEST "Ingo Weinhold" <bonefish@xxxxxxxxxxxxxxx> wrote: > On Sat, 02 Aug 2003 00:33:09 +0200 CEST "Axel Dörfler" <axeld@pinc- > software.de> wrote: > > At least, I don't see an urgent need to do it now, just to have it in > > there. > > I guess, you're right. This is okay with me too. StyledEdit as I have it committed right now doesn't require the OBOS libtextencoding.so, which will fail to build right now due to the missing implementations for the iconv functions. If an individual user wants to experiment with the OBOS implementation of this library, perhaps by exposing additional encodings, then they can do that by downloading libiconv from bebits and doing so. (Devs: feel free to contact me if this applies to you. :-P) > > Anyway, if we don't have a special directory for libtextencoding.so, > > but use kits/support for that, I would put iconv in that directory as > > well if it is to be part of libtextencoding, or kernel/libroot/posix > > if > > it's part of libroot.so, and libs/ if it's going to be an extra > > library > > - even if I would really dislike the latter, for the reasons > > explained > > earlier. > > Having it in a separate dir should make incorporation of updates > easier, I suppose. Haven't had a look at the library yet, though (if > it's only one file, then it wouldn't really be worth it ;-). > > CU, Ingo > The iconv library is expressed through about 200 source files. I have already made a version of it builds using jam and is located in the current/src/libs/iconv directory. (I have not checked this in. :-) ) As part of doing that I realized that I should really run "configure" for libiconv against OBOS instead of against beos, because of things like iconv.h. So, I tried to do that, which was less than straightforward, as you may imagine. During that process I mucked with a few posix headers in various ways. I also added a few and I am in the midst of ensuring that I haven't done damage to the build in general. So far my most troublesome change seems to be removing OS.h from posix/stat.h. This caused a problem in fs_interface.h that I fixed by including OS.h where it seemed appropriate. It also seems to have caused various problems elsewhere where uint32 and friends are no longer defined. There are miscellaneous other bits of code that inadvertently took advantage of this include. IMHO OS.h defines too many extra things to be included from a common posix header like stat.h. After I get all my changes working to my satisfaction I will post them to the list for public comment before I commit them due to the broad impact that header changes has. Feelings, thoughts, or opinions? Andrew