On 2014-05-05 22:48, Gilles Chanteperdrix wrote: > On 04/29/2014 05:39 PM, Jan Kiszka wrote: >> On 2014-04-25 14:00, Gilles Chanteperdrix wrote: >>> On 04/25/2014 12:40 PM, Jan Kiszka wrote: >>>> On 2014-04-05 14:14, Gilles Chanteperdrix wrote: >>>>> On 03/07/2014 09:19 PM, Gilles Chanteperdrix wrote: >>>>>> >>>>>> Hi, >>>>>> >>>>>> here comes a second attempt at introducing a file descriptor support for >>>>>> other purposes than RTDM drivers. >>>>>> >>>>>> This time, the file descriptors are called rtdm_fd and are part of the >>>>>> RTDM API, but can be used by the POSIX personality. The actual RB-tree >>>>>> where they are stored is part of the xnsys_ppd structure, as this way it >>>>>> can be used by one of the RTDM and POSIX personalities, even if the >>>>>> current process is not bound to the other personality. >>>>>> >>>>>> I have posted the patches to the timerbench, switchtest, rtdm, xddp, >>>>>> iddp and bufp drivers to allow seeing the API changes. >>>>> >>>>> Hi Jan, >>>>> >>>>> ping? >>>>> >>>>> I have a pile of code needing to be rebased on these patches, I would >>>>> very much like to get all this merged before it bitrots too much. >>>> >>>> Finally looking into them now. Do you happen to have them in git >>>> somewhere? Or what was the revision they once applied to? >>> >>> Branch for-forge-rtdm-rework in xenomai-gch.git. >> >> Thanks, worked through this now. I'm fine with the general approach. We >> should prepare it for merge. >> >> Some minor things I stumbled over while reading: >> >> - rtdm_context_user_p (I personally still prefer something like >> "is_user") and rtdm_context_device (to_device?) are apparently public >> APIs and, thus, need documentation >> - there are still things (arguments, functions etc.) called "context" >> that are of type rtdm_fd, thus should rather be called "fd", no? > > Hi Jan, > > I have started cleaning up the patches, however, about the last remark, > do you propose to rename rtdm_private_to_context to rtdm_private_to_fd > for instance? I was thinking that rtdm_fd was the new context, so we > could keep the context everywhere? Keeping the old API names would limit > the need for wrappers.
We can still keep (undocumented) legacy wrappers for a while, but it would be better to call the new API after what it is really translating. Jan -- Siemens AG, Corporate Technology, CT RTC ITP SES-DE Corporate Competence Center Embedded Linux _______________________________________________ Xenomai mailing list Xenomai@xenomai.org http://www.xenomai.org/mailman/listinfo/xenomai
