Hi, When I started adding the registry support for pSOS skin I first looked at the vxworks implementation. Then I implemented similar functionality in a similar manner for pSOS. So I make use of the registry code. These changes work fine.
However when adding some more changes where I wanted to make use of the "fuse_file_info->fh" data field I first noticed an issue. The fh data field is only supported in Fuse 2.5 and beyond. As we use Fuse 2.9.2 this should work but it did not. I first suspected the FUSE_USE_VERSION but this is (as you mention also) already defined at 26 for copperplate. To be sure that I did not do something wrong in my adjusted code, I just reverted my changes and build xenomai-forge as is with registry enabled. Even then when running a pSOS test application it seems to be using the fuse_main_real_compat22() function (I added some debug strings in the Fuse code). So I am a rather clueless now about what is causing this issue. I found a topic about a similar issue (not with Xenomai) on the internet where they assume linking issues (Binutils?) of the symbolic versioning. ( http://sourceforge.net/mailarchive/forum.php?thread_name=9e0cf0bf0711211110l5ab3cd0bpfcd624fde8efb928%40mail.gmail.com&forum_name=fuse-devel ) Not sure if the issues are related.... Regards, Kim 2013/10/8 Philippe Gerum <r...@xenomai.org> > On 10/08/2013 10:54 AM, Ronny Meeus wrote: > >> On Tue, Oct 1, 2013 at 12:13 PM, K. De Mey <kim.de...@gmail.com> wrote: >> >> Hello everybody, >>> >>> We use xenomai-forge for a project. I however came across a problem when >>> adding the "--enabled-registry" option. >>> It seems that the FUSE functions (for example fuse_main()) that are being >>> used are the compat functions of version 2.2 (fuse_main_real_compat22()). >>> I found this strange as I have FUSE_USE_VERSION defined at 26. >>> >>> I suspect it has something to do with the symbol versioning in FUSE. >>> I believe so as when I change the version 22 ".symver" mapping in FUSE to >>> use the newer compat functions then that works. (or removing all the >>> symver >>> completely works also) >>> This is obviously not the way to make it work, however I don't know how >>> to >>> do it properly. >>> I know this might be a FUSE issue and not necessary a Xenomai issue but I >>> can't seem to get it working. >>> I must say I know little to nothing of symbol versioning. >>> >>> As test I also created a stand alone FUSE application (no Xenomai >>> involved) >>> and then it does use the functions of the latest version. >>> So to be sure, I would like to know if there is something more in Xenomai >>> that sets the "compat" version of FUSE to 2.2? Am I missing something >>> ...? >>> >>> By the way, the actual version of FUSE that we use is 2.9.2 and the >>> version >>> of Xenomai-forge is also quite recent. >>> >>> I hope my question is a clear enough, else I can of course add more >>> information. >>> >>> Thanks! >>> >>> Regards, >>> Kim >>> ______________________________**_________________ >>> Xenomai mailing list >>> Xenomai@xenomai.org >>> http://www.xenomai.org/**mailman/listinfo/xenomai<http://www.xenomai.org/mailman/listinfo/xenomai> >>> >>> >> Hello guys, >> >> any feedback on the questions above? >> >> Kim is working on the introduction of the full registry support for the >> pSOS skin in forge. >> I think this is a major step to improve the debugability of applications >> using this skin. >> >> > Mentioning that you wanted to work on the registry support for a real-time > API gives the information which triggers an answer, since I could not > figure out how I could help Xenomai-wise, beyond the fact that we do set > FUSE_USE_VERSION appropriately and it does not seem to work for you > although it should. > > However, we set this definition only for building copperplate and the > registry daemon, since there is no reason to call any fuse library services > outside of this scope. So, if you intend to provide the registry support > for pSOS, please use the abstract registry infrastructure exposed by the > copperplate library (see vxworks for a -terse- example). If you think it's > not sufficient, then please submit some improvements to the abstract > interface, but don't call fuse services directly from the emulator code. > > I suspect the latter is the reason why you have the legacy fuse interface > referred to by your additional code to the pSOS emulator, since the pSOS > makefile does not (and should not) define FUSE_USE_VERSION. > > -- > Philippe. > _______________________________________________ Xenomai mailing list Xenomai@xenomai.org http://www.xenomai.org/mailman/listinfo/xenomai
