Hi folks, as we recently discovered that we need a new release, I finally managed to free up a whole evening (!) to add a few convenience changes I wanted to be in before we scare away new users by seemingly nonworking things. Here is what I did and a few suggestions what you all could do, apart from checking that my changes don't break anything. I'm about to commit them - should be in CVS in a few hours. O.K. - the changes and suggestions: > > What we should do though is to sit down and simply test > > as much of targets and functionality we can. > O.K. - I'll try a complete recompile/reinstall ASAP. +++ Works for me. ??? List: Please also check complete reinstall. > > The structs, checkmode, consistency and findleaks programs are a good start. --- Haven't done much on that yet. List: Please check that on all systems --- and targets you can get a hold of. Please report any inconsistencies. > >> * Remove _ggi_malloc and friends, use true error handling. > > Too much work to do before release. > Probably. A quick grep shows, that it isn't used much in the core, but quite > some in the targets. --- Pending. If someone with insight into the LibGGI internals has some free --- time - get at it. > >> * Fix the LibGGI internals so that error messages are only printed for > >> > required sublibs. > > Error messages for completely optional sublibs like > > default/fbdev/kgi/genkgi.so are confusing and annoying to users. Also > > we have some cases where a sublib probes for both devfs and old-school > > device nodes and only mentions the devfs-style name in the error > > message, which is also confusing. > Yes. +++ Fixed the devfs-style name stuff for fbdev target. Errors come in the +++ wrong order compared to the probe, but I didn't want to risk fprintf +++ changing the errno variable. ??? Folks: Please check that it works and doesn't cause more confusion. --- Haven't touched the other issues yet. Anyone ? BTW: Don't remove them - --- just make them DEBUG instead. > >> * Make input-linux-mouse parse /etc/XF86Config. > > This should be done, as most people have X configured properly, > > while few have a properly configured SVGAlib. > O.K. - I can take a stab at that one. However note, that it can be in quite > some places. RedHat puts it in /etc/X11/XF86Config, and seemingly SuSe even > reads ~/XF86Config or ./XF86Config - at least for root (GRRR !). +++ O.K. - done. Hope that works right. On my XGGI setup, +++ XF86Config is just "decoration". ??? Please test that it works. Note, that it doesn't get used, if the ??? device can be guessed by looking at /dev/mouse ... ??? I now look in /etc/X11/XF86Config and /etc/XF86Config. Anyplace else I ??? need to look for ? Just want to know about cases that aren't symlinked ??? to one of the already mentioned files. > >> * Make input-linux-mouse ignore config files with only comments. > > Shouldn't be too hard, and would make it possible to install an > > example config file. > I could take care of that while I'm at it. +++ Done. Files will be ignored, as long as they don't sepcify a +++ protocol. ??? Please Test --- Someone please add a template for $(prefix)/etc/ggi/input/linux-mouse --- with only explanatory comments in there and add it to the make install --- stage. The file should be ignored until it is edited to reflect the --- configuration. > >> * Make input-mouse print a warning when too many packets are bogus. > > Might be useful, but no showstopper. > Yes. Just to be friendly to lusers ... --- pending. Anyone ? CU, ANdy -- = Andreas Beck | Email : <[EMAIL PROTECTED]> =